STÖR Community, Integration, and Stability in Multinational Yugoslavia Steven L. Burg, Michael L. Berbaum The American Political Science Review, Volume 83, Issue 2 (Jun., 1989), 535-554. Stable URL: http://links.jstor.org/sici?sici=0003-0554%28198906%2983%3A2%3C535%3ACIASIM%3E2.0.CO%3B2-U Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://uk.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. The American Political Science Review is published by American Political Science Association. Please contact the publisher for further permissions regarding the use of this work. Publisher contact information may be obtained at http://uk.j stor.org/journals/apsa.html. The American Political Science Review ©1989 American Political Science Association JSTOR and the JSTOR logo are trademarks of JSTOR, and are Registered in the U.S. Patent and Trademark Office. For more information on JSTOR contactjstor@mimas.ac.uk. ©2002 JSTOR http://uk.jstor.org/ Tue Sep 10 11:57:27 2002 COMMUNITY, INTEGRATION, AND STABILITY IN MULTINATIONAL YUGOSLAVIA STEVEN L. BURG MICHAEL L. BERBAUM Brandeis University 1 n 1981 a higher proportion of the multinational population of Yugoslavia chose to declare "Yugoslav" in place of an ethnic identity in response to the census question on nationality than ever before. We present arguments to support the interpretation of Yugoslav identity as evidence of shared political identity, and carry out an analysis of aggregate data on the level of social and material development, political socialization, and interethnic contact in the country's nearly five hundred counties to discover the sources of that identity. We find that with certain important regional variations, Yugoslav identity seems to be the product of interethnic contact and higher education rather than the level of material well-being. These findings support an interpretation of Yugoslav identity as evidence of diffuse support for the existence of a shared political community and suggest both the sources and vulnerabilities of Yugoslav stability, L he stability of the communist political order in Yugoslavia rests in large part on the party's ability to maintain social and political peace among the nationalities. Interna-tionality conflict and nationalist unrest have challenged that stability in the past and constitute potential bases for such challenges in the future. Yet between the 1971 and 1981 censuses in Yugoslavia, the number of individuals who declared Yugoslav identity increased 41/2-fold/ from 273,077 to 1,219,024 persons, or from 1.3% to 5.4% of the population. The country is divided into almost five hundred counties with a median size of 446 square kilometers and a median population of 29,329. In many of those counties more than 10% of the population declared Yugoslav identity. In response to a more recent survey conducted by Yugoslav sociologists (Here 1987), some 15% of Yugoslav youth declared Yugoslav identity and some 36% expressed a "preference" for declaring it in place of their ethnic identity, suggesting that the tendency to declare Yugoslav identity has continued to spread. The increase in the number of "Yugoslavs" in the 1981 census far exceeded the expected natural increment of this group and can be attributed almost entirely to the declaration of this identity by individuals who had declared another ethnic identity in 1971 (see Appendix A). In some areas 'Yugoslavs" increased from near 0 in 1971 to over 10% in 1981. In other areas the proportion of 'Yugoslavs" has increased to over 15% of the population. In a few places it is now as much as 25%. Given the ethnic, social, demographic, and economic diversity of these counties, the declaration of Yugoslav identity by such proportions of their populations constituted a remarkable assertion of shared political identity, or "sense AMERICAN POLITICAL SCIENCE REVIEW VOLUME 83 NO. 2 JUNE 1989 American Political Science Review Vol. 83 of community." It is a clear indication of increased political integration and, given the particular meaning of such a declaration in the Yugoslav political setting, also implied increased levels of diffuse support for the regime. We present arguments to support the interpretation of Yugoslav identity as an indicator of political integration and diffuse regime support. We attempt, in turn, to explain the level of Yugoslav identification in terms of four, broad, social processes often associated with social mobilization and political integration: (1) material development, (2) the delivery of social services, (3) political socialization, and (4) intergroup contact. By doing so, we hope to be able to discover the nature and sources of political stability in Yugoslavia. Community, Integration, and Stability in Multinational States Cultural diversity, or multinationality, is widely viewed in the comparative politics literature as an impediment to integration at the mass level and therefore as a threat to political stability. In studies of single states the level of political integration at the mass level is often defined in terms suggested by Karl Deutsch, that is, in terms of the extent to which "a sense of community" has been achieved in the population (1953, 1957). The emergence of such a sense of community among the population of a multinational state need not necessarily be accompanied by the assimilation of cultural or national minorities, or the "homogenization" of society. As Amitai Etzioni suggests, even in an integrated political community "identification with the community is necessary only in political matters. Identification in other areas... might be less or more encompassing" (1965, 5). In effect, the process of integration in a multinational state can be viewed as the emergence of what Morris Janowitz calls "civic consciousness," or "positive and meaningful attachments" to the political community. It is, in Janowitz's terms, "the process by which national attachments and obligations are molded into the search for supranational citizenship" (Janowitz 1983, x-xi). Some scholars argue that political integration in the multinational state is enhanced by interethnic contact. Deutsch's "communications theory" of national identity formation, for example, is based very largely on indicators of such contact (1953, 70-79). More recently, Arend Lijphart and other theorists of the "con-sociationalism" approach have contested this view. They argue that integration is enhanced by the isolation of ethnic communities at the mass level, with contact reserved for ethnic elites (Lijphart 1975, 1977; McRae 1974). Empirical studies of integration have been handicapped by the difficulty of defining a reliable measure or indicator of the "sense of community" in the population. Deutsch's own work simply asserts that broad, socioeconomic developmental processes, subsumed under the rubric social mobilization may result in "integration" (1953, 1961). Essentially conceptual in its focus, it offers little empirical evidence to link these processes to changes in individual attitudes or behavior in the direction of an increasing "sense of community." Michael Hechter attempts to provide such evidence in his more recent study of the political economy of multiethnic integration in the United Kingdom (1975). Using county-level aggregate data of the type Deutsch suggests might be useful for the study of integration and nation building, Hechter explores the relationships among development (primarily industrialization), and persistence of ethnic (Celtic) identification, and support for the multinational regime, or political "community." Most contemporary research on political integration and regime support, however, has been based on the analysis of survey-generated data. Gary Bertsch, for 536 Nationality in Yugoslavia example, carried out a survey-based study of the "community-building process" in Yugoslavia in the late 1960s. He finds that social mobilization gave rise to a shift from traditional to modern values and, thereby, from particularistic to universal-istic patterns of identification. "In a complex multinational state such as the Yugoslav," he argues, "mobilization, modernism, and universalism are likely to engender the behaviors on which communities are based" (1972, 444). More recently, Allan Kornberg and colleagues have used survey data to explore the sources of "support for community" and "regime support" in the regions of Canada as a means of assessing the capacity of the Canadian regime to maintain its "integrity" (Kornberg, Clarke, and LeDuc 1978; Kornberg, Clarke, and Stewart 1980). They find that the level of political partisanship and intensity of party identification—factors that one might also interpret as indicators of mobilization—were powerful predictors of the level of support for community in Canada. Evidence of the behavioral consequences of varying levels of support has been provided in recent studies by Müller and his colleagues. They establish the existence of an inverse relationship between regime support on the one hand and anti-regime or "aggressive" political behavior on the other (Müller 1979; Muller and Jukam 1977; Muller, Jukam, and Seligson 1982; Müller and Williams 1980). Midler's research generates important empirical evidence that reinforces David Easton's earlier distinction between specific and diffuse support. Easton argues that specific support arises out of "the satisfactions that members of a system feel they obtain from the perceived outputs and performance of the political authorities." But, he continues, "it is only indirectly relevant, if at all, to the input of support for the regime or political community." Diffuse support, on the other hand, represents a "reservoir of favorable attitudes or good will that helps members to accept or tolerate outputs to which they are opposed or the effects of which they see as damaging to their wants." It is produced, Easton argues, by "childhood and continuing adult socialization" and may be manifest "as a sense of we-feeling, common consciousness, or group identification" (1975, 437, 444-45). Müller demonstrates that the level of diffuse support is far more powerfully related to aggressive political behavior than is the level of specific support. And this suggests that the level of diffuse support, manifest in the "sense of community," is most relevant to the study of political integration—and ultimately political stability—in multinational regimes. Measuring "Community" and "Support" Reliance on survey-generated data to develop an indicator of "community" in Yugoslavia is problematic. Indigenous scholars have only recently returned to large-scale survey research, and permission for foreign scholars to conduct such research may still be years away. Moreover, such research has until now proven unable to determine the causal direction of the relationship between objective social and economic conditions and therefore also of regime policy "outputs" on the one hand and the level of regime support in the population on the other (Müller and Williams 1980). Developing an acceptable indicator of political community has also proven problematic. Kornberg relies on a direct question, using a "thermometer scale" technique to record the level of support. Müller developed a set of attitudinal questions that, when scaled, seemed effectively to measure "system affect," or political support for the regime. Research based on aggregate data focuses on electoral behavior for this purpose. (Hechter, for ex- 537 American Political Science Review Vol. 83 ample, uses aggregate electoral support for the British Conservative party as his indicator). But the utility of doing so in a single-party state with a noncompetitive electoral system appears limited. Cohen and Warwick (1983), for example, attempt to circumvent this problem by using nonvoting in Yugoslav elections.1 Data that can be used as an alternative indicator of the "sense of community" or level of "diffuse support" for the existing political eommunity is, however, readily available in the Yugoslav case. It is census data on the declaration of Yugoslav identity in response to the open-ended question, "What is your nationality?" "Yugoslav" Identity and Political Community "Yugoslav" is not an easy answer to give in response to the census question on nationality. After a brief attempt to encourage "Yugoslavism" in the late 1950s, the official policy of the regime on this issue quickly became, if not anti- Yugoslav," at the very least pronational. In 1958 the central leadership had advanced a formal theory of nationalism and adopted a party program, both of which called for the emergence of a supranational, Yugoslav identity. But with the devolution of increasing power to the leaderships of the ethnically based republics and provinces in the early 1960s, this attempt to encourage 'Yugoslavism" was abandoned (Shoup 1968, 201-11). Since 1966 Yugoslav leaders have pursued a "consociational" strategy for controlling ethnic conflict. They have deliberately fostered segmentation of the social, economic, and political systems along ethnic and regional cleavages, reserving cross-cleavage interaction for ruling party elites (Burg 1983a). Since then, regional leaderships have seen any effort to encourage Yugoslavism as an attempt to undermine their respective power bases and to encourage support for a more cen- tralized political order. Consequently, they have opposed any official attempt to foster the emergence of an identity that might shift the focus of popular loyalties from the regions to the federal center. This reluctance formally to encourage the emergence of Yugoslav identity was reflected in the official instructions to census-takers concerning answers to the nationality question. In 1981 they stated that "if a citizen wishes to record Yugoslav as an answer to this question, the census taker is obliged to record even that answer, although by this the citizen does not declare membership in a nation or nationality" (Savezni Závod za Statistiku 1981, 38; emphasis added). Over time the meaning of Yugoslav identity appears to have changed. Until 1961 the category was used by the traditionally Islamic but ethnically Slavic population—for whom no category had yet been provided—to avoid having to declare either Serbian or Croatian identity. For the 1961 census the category "Muslim in the ethnic sense" was introduced. Nonetheless, significant numbers of Muslims appear to have continued to declare Yugoslav identity. Only in 1971, with the introduction of the category 'Muslim in the national sense," was the response 'Yugoslav" divorced from its past association with the Muslims (Burg 1983b). The growing sensitivity of the national question in the late 1960s led to explicit discussions of the meaning of Yugoslavism. As the 1971 census drew closer, the meaning of declaring Yugoslav identity became the subject of polemical debates. These appear to have resulted in the equation of Yugoslavism with socialist patriotism and the classification of Yugoslav as a political rather than ethnic identity (Bakic 1985). Moreover, at the time the 1971 census was being taken, internationally conflict resulting from the rise of assertive na-< tionalisms had already reached crisis proportions (Burg 1983a; Rusinow 1977). The most serious challenge to the legitimacy of the multinational community 538 Nationality in Yugoslavia was posed by Croatian student leaders and cultural elites who had mobilized nationalist sentiments and organized a popular, alternative political movement in that republic. Mobilized nationalisms in other regions also threatened to destroy the political community, but none represented as serious a threat as Croatian nationalism. Centered in the republic s universities, cultural institutions, and mass media, the Croatian nationalist movement quickly gained mass support. Its leaders advanced increasingly radical demands, and even came to challenge the communist party leadership for control over the republic. In late 1971 and 1972 this and the other regional political movements challenging the existing political order were forcibly suppressed, and their leadership arrested, purged, imprisoned, or forced out of public life. But their impact on Yugoslav political development was both powerful and long-lasting: they strengthened an already-existing tendency toward extensive devolution of political power and authority to the republics and provinces at the expense of the federation, a tendency that remains a powerful force in Yugoslav politics even today (Burg 1986). Those who declared "Yugoslav" rather than an ethnic identity in 1971 were deciding in favor of an option that clearly was contrary to the overwhelmingly centrifugal tendencies of the moment. At the very least, the declaration of Yugoslav identity in 1971 can be interpreted as an indication of the conscious rejection of a fissiparous, or exclusivist, orientation in favor of a more integrative, or universalist, one. By 1981, however, the definition of Yugoslav as a political identity appears to have become more clearly established in both the popular press and more scholarly works (Dugandzija 1984; Matvejevic 1983). From this perspective, Yugoslav identity can be interpreted as an expression of what Janowitz has called "civic consciousness" or what one Yugoslav political scientist has called a "transnational" sense of "belonging to the society, the community" (Miric 1984, 109). Some behavioral evidence of an association between "Yugoslav" identity and support for the overarching political community can be found in the overrepresen-tation of "Yugoslavs" in the Party. (All party data are from Cvjeticanin 1984.) While only 1.3% of the population in 1971, "Yugoslavs" constituted 3.9% of the Party. In 1981 they constituted 6.5% of the Party (7.1% in 1982). The proportion of "Yugoslavs" in each of the regional party organizations for which they were reported increased steadily from 1972 to 1982, despite the overall growth of these parties. By 1982, they were overrepresent-ed in the party organizations of Slovenia, Croatia, Bosnia, and Serbia. More importantly, 'Yugoslavs" were also overrepre-sented in those sectors of the party where we might expect commitment to the regime to be highest: the military and federal party organizations. In 1981 "Yugoslavs" constituted 8.0% of the military party organization, which comprises primarily the officer corps. In 1982 they were 8.8%. In the federal party organization, which comprises all party members employed in federal institutions and organizations, "Yugoslavs" constituted 19.5% of the total in 1981 and 18.0% in 1982. While this suggests a strong association between Yugoslav identity and support for the shared political community, it is important to note that the actual strength of that relationship may be even greater. Regional leaderships exercise considerable influence over appointments to the federal apparatus. This represents a strong incentive to declare the titular national identity of one's home region, and may suppress the declaration of Yugoslav identity. If we assume that membership in the party reflects support for the multinational political community, these data are consistent with the interpretation of 539 American Political Science Review Vol. 83 Yugoslav identity as evidence of such support. However, the declaration of Yugoslav identity does not necessarily indicate specific support. It may indicate only diffuse support for the idea of a shared, multinational, political community. As Easton has suggested, such support does not preclude opposition to the incumbent leadership and its policies. An Alternative Hypothesis The dramatic increase in the number of "Yugoslavs" in 1981 led to renewed denials of, and opposition to, any official effort to encourage the emergence of a Yugoslav ethnic identity. It also led some Yugoslav observers to suggest explanations for the increase having little to do with popular support for the regime.2 Foremost among these was the attempt to attribute the declaration of Yugoslav identity to the effect, on both the marriage partners and their offspring, of mixed marriages. Were county-level data on the rate of mixed marriages to become available, it would hardly be surprising to find a strong correlation between such rates and the declaration of Yugoslav identity. The selection of an ethnically distinct marriage partner has long been recognized in studies of ethnocentrism and ethnic distance as the strongest evidence of ethnic tolerance or openness, surely a precondition to the emergence of a "sense of community" in a multiethnic setting. In the absence of cross-time data on the marital status and nationality declarations of individuals, however, it is impossible to determine directly whether intermarriage results in a shift from ethnic to Yugoslav identity. However, the effect of ethnic intermarriage on the declaration of Yugoslav identity by offspring can be subjected to such a test. Since parents report the national identity of their children under 15 years of age, we used census data on age to create a "generation" of individuals who attained the age of 15 in the intercensal period and were declaring their identity on their own for the first time in 1981. This generation included the offspring of mixed marriages concluded in the 1960s. If the increase in the number of 'Yugoslavs" reflected an increase in the number of individuals who were the offspring of mixed marriages, we would expect strong positive correlations between the proportion of the population in this generation and the proportion that declared Yugoslav identity. Yet, as will be shown, no such relationship could be established. A more refined interpretation of the relationship between mixed marriages and Yugoslav identity was offered by the results of a recent survey of attitudes and values among youths, conducted by a group of Yugoslav sociologists. These researchers asked their respondents to distinguish between their "objective" nationality and their "preferred" national identification. While 34% of the offspring of mixed marriages declared Yugoslav identity, approximately 14% of the children of homogeneous marriages also did so. And whereas 51% of the offspring of mixed marriages "preferred" to identify as 'Yugoslav," approximately 35% of the children of homogeneous marriages also "preferred" to do so (Here 1987; calculations based on his Table 2, p. 11). A 1982 survey of youths from all over Yugoslavia participating in political summer camps in Croatia yielded similar results: 38.4% of the 1,729 respondents declared Yugoslav identity, and 51% reported they "felt themselves" to be "just Yugoslav." While 52.6% of the offspring of mixed marriages declared Yugoslav identity, 39.4% of the children of homogenous marriages also did so. Moreover, the children of mixed marriages comprised only 16.3% of these politically mobilized youth, clearly not a large-enough group to account for the high proportion of "Yugoslavs" (Obradovic 1984, 140-44). These findings suggest that mixed mar- 540 Nationality in Yugoslavia riages may be intensifying processes that are taking place more broadly in Yugoslav society, or even contributing to re-identification independently. But it is very unlikely that mixed marriages alone account for the scope of Yugoslav identification. It is far more likely that both Yugoslav identity and mixed marriages are independently associated with ethnic interaction (cf. Obradovic 1984, 250). The Correlates of Yugoslav Identity The Research Design Using the proportion of 'Yugoslavs" in the population of each county as an indicator of support for a shared political community and as the dependent variable in our analyses, we attempt to explain the level of support in terms of four broad social processes: (1) material development, as measured by per capita income (which is very strongly and positively correlated with such other indicators as per capita gross material production and industrial employment and negatively correlated with such indicators of underdevelopment as the proportion of peasants in the population) and by the availability of expendable income (i.e., the extent of television ownership) in each county; (2) the delivery of social services, as measured by the number of doctors and dentists per thousand population (which is highly correlated with the number of individuals employed in education, culture, health, and other social services and the number of hospital beds per thousand population); (3) political socialization, as measured by the proportion of the population in each county with higher education and by newspaper circulation (television ownership is relevant here, as well); (4) intergroup contact, as measured by both the degree of ethnic heterogeneity, or "fragmentation," of the county population;3 and (for 1981) the proportion of the population that had already declared Yugoslav identity in 1971. In addition, we explored the effect of generational differences in the population on the extent of Yugoslav identification. We did this in terms of "political generations," defined as age cohorts whose adolescent and young adult years came during (1) World War II and the revolution, (2) the postwar de-Stalinization, (3) the political liberalization of the late 50s and early 60s, (4) the rapid regionalization and the rise of ethnic nationalisms in the late sixties and early seventies, and (5) the consolidation of the decentralized system in the later 1970s. The material and social development variables were intended to measure two dimensions of policy "outputs." If Yugoslav identity is primarily a manifestation of specific support, these variables can be expected to be powerful predictors of it. If, on the other hand, Yugoslav identity is a manifestation of the level of diffuse support, we should expect the measures of political socialization, intergroup contact, and generational experiences to be more powerful predictors of the proportion of the population declaring Yugoslav identity. The proportion of 'Yugoslavs" in any county is the product of both the personal characteristics of individuals, such as life history and personality, and the social and economic characteristics of both individuals and the county setting itself. No personalistic variables were available in our census data and in any event the distribution of variations in these characteristics should not vary much across county populations. Our analysis was conducted entirely on the basis of aggregate data about the social and economic characteristics of county populations. Our intention was to relate the occurrence of a key political behavior—the declaration of Yugoslav identity—to fundamental variations in the characteristics of these social collectivities. Aggregation biases can arise through 541 Table 1. Median County Values on Predictor Variables, 1981 en All Yugoslavia Regions South North Variables Kosovo Bosnia Serbia Montenegro Macedonia Vojvodina Croatia Slovenia Material development Per capita income (dinars) TVs per thousand 15,928 166.8 4,697 40.0 12,241 118.7 12,429 143.1 13,573 115.6 13,880 118.1 17,015 219.3 20,146 208.7 34,310 255.1 Social services Doctors per thousand .9 A ,7 1.0 1.0 ,9 1.1 1.1 1.0 Socialization % with higher education Newspaper circulation per thousand 2.8 347.1 1.9 .9 2.6 164.2 2.4 149.8 4.0 532.1 1.9 296.7 3.2 499.3 3.3 592.0 3.6 2433.0 Contact Fragmentation % "Yugoslav" 1971 .28 ,42 .29 .03 .52 ,40 ,11 ,33 .40 .65 .22 .06 .53 1.32 .33 1.05 Ml .20 Generations (%) War De-Stalinization Liberalization Regionalization Consolidation Under age 15 18.8 13.3 11.9 15.2 16.1 22.2 8.7 7.2 9.1 12.4 19.8 42.0 12.0 12.1 11.5 15.1 21.1 26.3 21.6 16.1 12.0 15.4 14.0 20.0 16.3 11.2 11.8 14.7 18.7 25.2 14.0 10.7 11.6 15.9 18.9 26.9 22.2 15.1 12.6 15.3 14.2 19.2 22.7 14.3 11.8 14.5 15.0 20.1 18.9 12.6 12.1 16.0 16.0 22.8 Number of counties 486 22 104 107 20 30 44 103 56 Nationality in Yugoslavia omission from the model of important explanatory variables. Given the broad inclusion of aggregate-level variables from each of four theoretical viewpoints in our model, omitted variables in the present analyses were most likely to be of the per-sonalistic kind, that is, variables that distinguished individuals. These may reliably affect declaration of Yugoslav identity. But the bias introduced by their omission will be small if the basis of grouping was such that little or no correlation across the groups (counties) existed between these omitted personalistic variables and the other aggregate variables that were included (Hanushek, Jackson, and Kain 1974). Most of the personalistic variables that might affect the declaration of ethnic identity must be distributed randomly across the 486 counties, so the correlations that could produce bias was presumably near zero in magnitude. Of course, we cannot support a claim that this is true for all such variables; possibility of some aggregation bias in our coefficient estimates cannot be denied. On the whole, however, we consider that we were much closer to the best than to the worst case4 and further believe that our conclusions would be little altered by elimination of any remaining aggregation biases from our estimates. The consistently high R-squareds associated with our models support this conclusion. As the data displayed in Table 1 make clear, the social and economic characteristics of counties and their populations vary enormously across the regions, so that analysis at the all-Yugoslavia level alone is likely to obscure important regional variations in the statistical relationships among our variables. The country as a whole is divided between historically, culturally, and developmentally distinct northern and southern zones (Burks 1971). The republics of Slovenia and Croatia and the autonomous province of Vojvodina (a part of the republic of Serbia) constitute the relatively highly devel- oped north, whereas the republics of Bosnia and Hercegovina, Montenegro, and Macedonia, and the autonomous province of Kosovo (another part of Serbia) constitute the underdeveloped south. "Serbia proper," that is, Serbian territory apart from the autonomous provinces, can also be considered part of the less developed south. The magnitude of developmental differences between the north and south is reflected in the data presented in Table 1. The median per capita income in the counties of the most-developed northern region, Slovenia, for example, is six times that in the least-developed southern region, Kosovo. The median county values of almost all our other predictor variables are also characterized by similarly sharp regional differences. Even within the north and the south, however, the individual republics and provinces are characterized by distinct political histories and cultural development and by important differences in ethnonational composition and demographic trends. Such sharp regional differences must be taken into account in any analysis. We approached them in much the same way as Kornberg and his collaborators did in their study of public support for "community" and "regime" in the regions of Canada (Kornberg, Clarke, and Stewart 1980); that is, we treated them first as regionally distributed variations within a single pattern or set of processes by introducing a set of dummy variables for region into an all-Yugoslavia analysis. Second, we treated them as reflections of the fact that the republics and provinces are distinct places characterized by distinct processes. We therefore carried out eight separate but, for purposes of comparison, parallel analyses.5 The differing results produced by these two approaches are illustrated by the simple correlations between our predictor and dependent variables, displayed in Table 2. Note, for example, that the correlation between per capita income and the proportion of 543 Table 2. Pearson Correlations between Explanatory Variables and Proportion of "Yugoslavs" in the Population, 1981 All Yugoslavia Regions South North Variables Kosovo Bosnia Serbia Montenegro Macedonia Vojvodina Croatia Slovenia Material development Per capita income (dinars) TVs per thousand Social services Doctors per thousand Socialization % with higher education oj j Newspaper circulation per ^ I thousand Contact Fragmentation % "Yugoslav" 1971 Generations (%) War De-Stalinization Liberalization Regionalization Consolidation .25 .31 .36 .48 .12 .59 .80 .06 .20 .27 .28 .12 .56 .57 .56 .65 .72 .57 .38 .12 .35 .55 .74 J7 .66 .69 .47 .80 .28 .54 .26 -.02 .53 .50 .55 .70 .75 .82 .56 .99 .32 -.05 -.02 .19 .47 .37 -.09 .41 .27 .59 .20 .55 .54 .59 .20 .73 .09 -.42 -.07 -.74 .29 .08 .36 .23 .47 .45 .55 .23 -.03 .03 .17 -.14 .36 .29 .19 .41 .36 .33 .52 .29 .44 .43 .38 .67 .30 .25 .24 .38 .80 .80 .81 .82 .77 -.34 -.10 -.15 .29 .20 .23 .22 .12 .24 .31 .36 .69 .17 -.10 -.42 Number of cases 486 22 104 107 20 30 44 103 56 Nationality in Yugoslavia "Yugoslavs" was only .25 for the country as a whole, while it varied from .29 to .65 in the regions. The relationships in each region between our predictor and dependent variables are likely to differ in character, as well as in magnitude. The relationships between higher education and "Yugoslav" identity in Croatia, for example, is almost certain to be very different from that in the other regions. Both the magnitude and the character of these relationships can be expected to differ over time as well. Devolution of power and authority to the regions and their increasing control over public policies, for example, are likely to dilute the impact of policy outcomes on the degree of support for the overarching multinational political community. Thus, any attempt to explain the sources of Yugoslav identity in 1981 must take into account both interregional and cross-time differences in the relationships between variables. Findings Although it is impossible to measure the unique contributions of each explanatory variable to the declaration of Yugoslav identity, we developed an approximation of their relative importance by applying multiple linear regression analysis techniques to our data.6 For purposes of "intelligibility," we present the results of our regression analyses in the form of standardized regression coefficients (Converse and Pierce 1986, 958, n. 7). These are displayed in Table 3. Since our cases comprise all existing counties, rather then a sample of them, these should be treated as descriptive measures rather than inferential ones. Analysis at the all-Yugoslavia level produced strong results, accounting for about 72% of the variation in the proportion of "Yugoslavs" among the 486 counties. With all of the explanatory variables taken into account simultaneously, the proportion of 'Yugoslavs" in 1971 and fragmentation (our measure of the level of interethnic contact) were both strongly and positively associated with the declaration of Yugoslav identity in 1981. The proportions of the population in the de-Staliniza-tion generation, and with higher education were moderately associated with the declaration of Yugoslav identity. The extent of newspaper readership was moderately and negatively associated with "Yugoslavs." No other variable approached these in relative importance. Contrary to the hypothesized effect of interethnic marriage on the identification of offspring, there was essentially no association between the proportion of the population in the "consolidation" generation (comprising individuals declaring their identity on their own for the first time) and the proportion of "Yugoslavs" in the population. Similarly, neither per capita income in 1980 nor the extent of television ownership—both good indicators of material well-being—were strongly associated with regime support as manifested in the declaration of Yugoslav identity, contrary to expectations. By adding a set of dummy variables for region to this analysis (the all-Yugoslavia-plus-region column and the dummy-co-efficient-for-all-Yugoslavia-plus-region row), we produced even stronger results. The proportion of variation in 'Yugoslavs" accounted for by this model increased to about 77%. The addition of region also resulted in a sharp decline in the relative strengths of education and newspaper circulation as predictors of Yugoslav identity. The strength of the association between interethnic contact and "Yugoslavs," in contrast, remained essentially unchanged. And the strength of the proportion of "Yugoslavs" in 1971 as a predictor of "Yugoslavs" in 1981 actually increased. The regression coefficients for the dummy variables for region suggest that evert if we assume that the socioeconomic proc- 545 Table 3. Standardized Regression Coefficients for 1980-81 Variables All Yugoslavia All Yugoslavia + Region Regions South North Kosovo Bosnia Serbia Montenegro Macedonia Vojvodina Croatia Slovenia en Material development Per capita income (dinars) -.041 TVs per thousand .060 Social services Doctors per thousand -.035 Socialization % with higher education .114 Newspaper circulation per thousand -.103 Contact Fragmentation .325 % "Yugoslav" 1971 .599 Generations (%) War .083 De-Stalinization .142 Liberalization -.006 Regioňalization .000 Consolidation .070 R2 .73 Adjusted R2 .72 Number of cases 486 Dummy coefficient for all Yugoslavia + region — .027 .071 -.034 .017 -.076 .297 .615 .054 .082 .005 .020 -.093 .78 .77 486 .234 .243 .255 22 -.028 .165 .135 -.053 -.243 -1.413 .364 104 .075 107 -.120 .171 .126 .200 676 -.272 2.039 -.137 417 .119 -1.184 .026 065 .153 -.127 .019 568 .558 .254 1.154 169 -.094 .331 -.000 752 .192 -.070 .077 491 -.141 -.024 -.119 285 .125 .083 .020 064 -.411 -.369 .089 92 .83 .61 1.00 .81 .80 .57 .99 20 -.073 .452 -.066 .612 -.020 .350 .101 .187 -.200 -.679 .242 -.147 .64 .40 30 .095 -.204 .291 -.260 44 .186 .039 .054 .070 103 -.161 -.040 .018 .222 -.265 .564 .009 -.036 -.227 .034 .805 .538 .474 .143 .462 .047 .251 .013 .127 .087 .002 .025 -.061 .165 -.084 .086 .146 -.009 .247 .131 .77 .68 .83 .81 .84 .80 56 .166 flBase category for region. Nationality in Yugoslavia esses that produce "Yugoslavs" are the same everywhere, these processes will produce marginally smaller proportions of "Yugoslavs" in Montenegro, Macedonia, Slovenia, Serbia, Kosovo, and Vojvodina than in Croatia and will produce a marginally greater proportion of "Yugoslavs" in Bosnia. We explored these regional effects more closely by subjecting the data in each republic and province to separate but identical analyses. The results of identical regional regressions are displayed in the regions columns in Table 3. The most striking results of this approach were, first, that the pattern of associations varied dramatically across the regions, a result consistent with the vast economic, cultural, and other differences between them, noted above; and, second, that with the exceptions of Macedonia and Serbia, regional-level analysis offered even greater explanatory power. Note that in five of the eight regions we again accounted for more than three-quarters of the variation in our dependent variable. Our findings with respect to material development reflect the north-south division in Yugoslavia. Our indicator of material development—per capita income—was relatively moderately and positively associated with Yugoslav identity in Bosnia, Montenegro, Macedonia, and Kosovo. These four regions constitute most of the underdeveloped south of the country and are heavily dependent on the central government for developmental assistance. It is not surprising, therefore, that higher levels of development in these regions should be associated with higher levels of political support for the multinational community and, therefore, higher proportions of "Yugoslavs." Our indicator of social development—the number of doctors per thousand population — yielded mixed results in these regions, a finding consistent with the narrower interregional differences on this dimension reported in Table 1. The association between the proportion of "Yugoslavs" in 1971 and the proportion in 1981 was positive and relatively strong in every region except Macedonia and Serbia, where it was positive but relatively weak. The association between inter-ethnic contact and Yugoslav identity was relatively strong and positive only in Croatia and Macedonia. In Croatia the level of interethnic contact was, in fact, the strongest predictor of Yugoslav identity. Thus, except in Serbia, either the level of interethnic contact or the proportion of the population that had already declared Yugoslav identity in 1971 or both were among the strongest predictors of Yugoslav identity in 1981. These findings support the proposition that interethnic contact contributes to political integration. The strong positive association between the proportion of "Yugoslavs" in 1971 and the proportion in 1981 suggests that Yugoslav identity may be spread through a process of socialization. Indeed, consistent with this view, the association between the proportion of the population with higher education and the proportion of "Yugoslavs" was strongly positive in Serbia, Kosovo, and Slovenia. But it was strongly negative in Croatia—where nationalism has been strongest in the universities—and in Bosnia. Newspaper circulation, our other indicator of exposure to formal socialization processes, was positively associated with Yugoslav identity only in Bosnia, where the press tends to be more supportive of the central regime. Interregional differences in the nature of socialization processes and their relationship to Yugoslav identity were reflected in the relationships between the proportions of the population in each of the five "political generations" and the extent of Yugoslav identity. No single generation was consistently and strongly associated with Yugoslav identity across all regions. Nor was there a consistent pattern of increasing or decreasing association across 547 American Political Science Review Vol. 83 the generations. Surprisingly, in both Slovenia and Croatia the proportion of the population in the regionalization generation, which came to political maturity in a period characterized by the rapid decentralization of power and authority to the regions and by the rise of nationalism in both these republics, was positively and, in comparison to other explanatory variables, moderately associated with the declaration of 'Yugoslav" identity. Contrary to the hypothesis that mixed marriages concluded in earlier years might have resulted in an increased rate of declaration of "Yugoslav" identity among those declaring their nationality on their own for the first time, the consolidation generation was not consistently more strongly associated with Yugoslav identity than the older generations. In most regions the relationship of the consolidation generation to Yugoslav identity was negative and relatively strongly so in Bosnia. Also contrary to expectations, the oldest, war generation was more strongly associated with Yugoslav identity than younger generations only in Serbia. Overall, these findings suggest that the sources of Yugoslav identity and, by implication, the nature of support for the multinational community vary in important ways from region to region. Yugoslav identity is in part the product of material development, or policy "outputs," in the country's underdeveloped south. In these regions, therefore, support for community is at least partially specific in nature. But in almost every region, north and south, Yugoslav identity is the product of certain forms of socializing influences: interethnic contact, contact with already-declared "Yugoslavs," the completion of a higher education, or even formative political experiences. This indicates that the declaration of Yugoslav identity in 1981 was everywhere at least in part a reflection of diffuse support. Cross-time Comparisons In order to establish the cross-time validity of this analysis, we compared these results to results derived from an analysis of 1971 data identical to our 1981 data, with two important exceptions: first, newspaper circulation statistics were not available for 1970; second, there is no comparable measure of the number of "Yugoslavs" in the previous (1961) census.7 Results of our analysis of 1971 data paralleled our 1981 findings.8 Analysis at the all-Yugoslavia level indicated that in the period before the assertion of Yugoslav political identity by significant proportions of the population, the level of interethnic contact was the most powerful predictor of such identity. The proportion of the population with higher education was the next-most-powerful predictor. At the regional level the pattern of associations in our 1971 data between social and material conditions and the declaration of Yugoslav identity was similar to that in 1981; that is, social and material development were relatively unimportant in the developed north but more important in the less-developed south. The relationship between interethnic contact and Yugoslav identity also paralleled the 1981 findings; that is, contact was important in every region except Serbia. An important difference that emerged between our 1971 and 1981 results was the relationship between the proportions of the population with higher education and the proportion of "Yugoslavs." In the 1971 data, that relationship was moderately-to-strongly positive in all of the underdeveloped regions and in Vojvodina. It was moderately negative only in Slovenia, where a period of nationalist discontent had only recently come to an end. In Croatia, where nationalism was reaching its zenith, higher education and Yugoslav identity seemed unrelated in 1971. In our 548 Nationality in Yugoslavia 1981 data, however, this relationship was negative in three of the five underdeveloped regions and strongly negative in Croatia. These differences imply that the region-alization of the educational system that resulted from the devolution of the 1970s had eroded the socializing and integrative political effect of higher education in Yugoslavia. Indeed, our findings lend indirect support to the commonly held notion that a university education socializes one to Croatian nationalism in Croatia, tends to foster Bosnian Muslim ethnic and political identity in Bosnia, and contributes to the promulgation of Macedonian national identity in Macedonia. Only in Serbia has the completion of higher education remained strongly associated with the declaration of Yugoslav identity over time. And only in Slovenia has the relationship changed from negative to positive. Similarly, the regionalization of control over the mass media during the 1970s helps to explain the absence of any consistent pattern of associations between Yugoslav identity and newspaper circulation in 1980, Conclusions Where diffuse support for a regime is sizable or increasing among the mass population, the political order can be expected to remain stable even during a period of declining regime performance. In the conflictual, multinational context of contemporary Yugoslavia, the declaration of Yugoslav identity can be seen as evidence of the "sense of community" associated with such diffuse support; and the increase between 1971 and 1981 in the proportion of the population that declared it suggests one reason for the stability of the political order until now in the face of sharply declining regime performance since 1981 (Burg 1986), The unimportance of the level of material development as a predictor of popular support in the country's north and its relative weakness even in the country's south in the period 1971-81 may help to explain the surprising resilience of the regime in the face of the dramatic economic decline in the period since 1979. Our findings indicate that the diffuse support for the multinational community manifested in the declaration of Yugoslav identity was attributable in greater part to broad social processes that were largely independent of the level of development and can consequently be expected to continue to unfold for the foreseeable future: increasing ethnic interaction—especially the socializing influence of already-declared "Yugoslavs" on others—and, to a lesser extent, the spread of higher education. The declaration of Yugoslav identity appears to fall into the category of what Schelling calls 'critical mass phenomena" (1978, 91-102). In effect, the combination of higher education and interethnic contact induced individuals to declare Yugoslav identity in 1971, and the presence of these 'Yugoslavs" increased the rate at which others decided to do the same. These findings support the view that while ethnic segmentation and isolation may be conducive to the establishment of elite cooperation and a "consociational" political order, interethnic contact and interaction is conducive to the emergence of the "sense of community" that fosters political integration at the mass level. This implies that all other things being equal, the declaration of Yugoslav identity is likely to become more widespread, and the level of support for the multinational community is likely to increase. Some contemporary evidence suggests that this may in fact be the case. The post-census surveys of Yugoslav youth cited earlier reported rates of declaration of Yugoslav identity many times greater than that of the 1981 census and astounding levels of "preference" or "feeling" for 549 American Political Science Review Vol. 83 Yugoslav identity (Here 1987; Obradovic 1984). But the key factor here would appear to be the ability of the regime to hold "all other things equal." We have already noted that our findings suggest a decline in the contribution of both higher education and exposure to the mass media to the level of diffuse support in Yugoslavia. It is also important to remember that the absolute levels of economic and social development of almost all the countries improved in the period 1971—81, their relative levels remaining largely unchanged. Fragmentary but mounting evidence indicates that the sharp decline in economic conditions since 1981 has had a substantial negative effect on popular assessments of the government and is resulting in increased levels of pessimism in the population.9 Among intellectuals, students, and others, criticism of the regime's present incumbents and their policies and demands for democratization of the political order are increasing.10 Thus the regime is already experiencing an erosion of specific support. While short-term negative performance may not affect the level of diffuse support, sustained negative performance is likely to do so. If the Yugoslav leadership is to insulate diffuse support for the regime from such an erosion, it must first resolve the social, economic, and political problems that underlie the decline in economic performance and contribute to the erosion of specific support. But the findings reported here imply that in order to increase the level of diffuse support, the leadership must also revise its strategy for controlling ethnic conflict and devise solutions to break down ethnic isolation and permit increased levels of interethnic contact and to renew the formal socialization of regional populations to the Yugoslav idea. In the absence of such changes, it may be very difficult indeed to sustain diffuse support for a multinational political community in Yugoslavia. Appendix A Estimated 1981 Population Deficits and Surpluses, by Group A B Projected C D 1981 Deficit E 1971 Natural Increment 1981 or Surplus Das % Group Population 1971-80 Population (C minus [A + B]) of (A + B) Montenegrins 508,843 60,070 579,043 10,110 1.8 Croats 4,526,782 262,126 4,428,043 -360,865 -7.5 Macedonians 1,194,784 139,490 1,341,598 7,324 .5 Muslims 1,729,932 316,001 1,999,890 -49,043 -2.4 Slovenes 1,678,032 90,989 1,753,571 -15,450 -.9 Serbs 8,143,246 539,865 8,140,507 -537,126 -6.2 Albanians 1,309,523 441,780 1,730,878 -20,426 -1.2 Bulgarians 58,627 2,512 36,189 -24,950 -40.8 Hungarians 477,374 -7,568 426,867 -72,939 -15.5 Romanians 58,570 -1,354 54,955 -2,261 -4.0 Romi/Gypsies 78,485 25,289 168,197 64,423 62.1 Slovaks 83,656 1,209 80,334 -4,531 -5.3 Turks 127,920 16,787 101,291 -43,416 -30.0 "Yugoslavs" 273,077 55,227 1,219,024 890,720 271.3 Total 20,522,972 1,916,606 22,427,585 -11,933 -.1 Source: Ruza Petrovic, "The National Composition of the Population/' Yugoslav Survey 24, no. 3 (August 1983), Tbl. 3, p. 26; Štatistickí Eilten, no. 1295, p. 8. 550 Appendix B Correlations of Explanatory Variables with 1981 Population Density Variables Material Social Development Services Socialization Contact Generations (%) Newspaper Per TVs Doctors % Circulation % (n Capita per per Higher per "Yugoslav" M Regions Income Thousand Thousand Education Thousand Fragmentation 1971 War De-Stalin Liberalization Regjonalization Consolidation All Yugoslavia .32 .28 .46 .60 .82 .25 .26 -.18 -.02 .28 .32 -.07 Kosovo .37 .49 .46 .61 .51 .28 .35 -.63 -.43 .04 .18 -.15 Bosnia .47 .53 .55 .65 .51 .28 .70 -.34 -.04 .39 .50 -.13 1 Serbia .51 .56 .57 ,7S .93 .51 .70 -.18 .00 .32 .35 -.08 Montenegro .27 .66 .28 .61 .23 .75 .65 -.19 .21 .47 .65 -.50 Macedonia .19 -.17 .04 -.07 .82 .65 .71 .26 -.27 .07 .11 -.00 1 Vojvodina .76 .39 .70 .86 .80 .12 .49 -.57 -.39 .58 .69 -.02 Croatia .42 .34 .51 .60 .69 .26 .18 -.26 -.10 .39 .31 -.13 Slovenia .60 .48 .64 .65 .44 .60 .29 -.32 .40 .60 .53 -.22 American Political Science Review Vol. 83 Appendix C Census data, including age distribution, nationality, and educational levels of the county populations are taken from various volumes of the multivolume final census reports of the Federal Institute for Statistics (Savezni Závod za Statistiku). The 1971 census results were published by the institute under the series title Popis stanovnistva i stanova, 1971 [Census of the population and apartments, 1971] in Belgrade in 1974. The 1981 census results were published in Belgrade during 1983 and 1984 by the institute under the series title Popis stanovnistva, domacinstava, i stanovnistva i stanova, 1971 fCensus of tion, households, and apartments in the year 1981]. Some economic data, sectoral employment data, data on the number of doctors and dentists, and data on television ownership are reported in the annual statistical handbooks of the institute, entitle Popis stanovnistva, domacinstava, i [Statistical yearbook of the SFR of Yugoslavia], usually with a one- to three-year lag. English translations of the introductory materials, table headings, and labels are available for each of these volumes. Other economic data and newspaper circulation data for 1980 are published in the institute's periodical publication Štatistickí bilten [Statistical bulletin], which appears about 40 times a year and reports a wealth of social, political, economic, and cultural data. Each of these publications contains definitional materials for the data reported. Notes 1. It would seem to be far more promising to use aggregate data on membership in the ruling party for this purpose. However, even use of these data, were they available, would require careful specification of the social and political meaning of the decision to join the Party. For example, does it reflect primarily ''regime support" or an attempt to advance one's career prospects? Is party membership a mass or an elite phenomenon? In any event, we were unable to obtain such data in a statistically useful form. 2. See, for example, the series of exchanges in the following: Dusan Bilandzic's interview in Vjes-nik, 8 May 1982; the commentary, "Suspect Yugoslavs," in Politika, 13 May 1982; the summary of reactions to the Bilandzic interview in Politika, 22 May 1982; and the article by Bilandzic in Politika, 23 May 1982. For empirically based discussion, see the studies Vuskovic 1982 and Raic 1982 and the telephone survey reported in "Otkud 'plima jugo-slovena'?" [Whence "the tide of Yugoslavs"?], NIN, 23 May 1982, pp. 18-21. 3. The exact formula is n f 2 Fragmentation * 1 — L ( _ ) í-i N N ' where f{/N equals the proportion of individuals in the county in nationality group i (Rae and Taylor 1970). 4. Lichtman, for example, argues that "estimates of individual behavior from information pertaining to aggregate units are not necessarily less useful than estimates from data pertaining to smaller units or even to individuals." He goes on to suggest that "if aggregation tends to randomize independent variables, it can reduce specification error and facilitate the goal of portraying the independent effects of individual behavior. Moreover, historians will frequently be able to measure a significantly greater number of relevant variables at higher levels of aggregation. For instance, a much wider variety of statistical measures is generally available for counties than for voting precincts. As a result, estimates derived from county-level data may be more useful than estimates derived from precinct-level data. A well-specified model at a higher level of aggregation may yield more precise and reliable estimates of the direct influence of independent variables than a poorly specified model at a lower level of aggregation" (Lichtman 1974, 432). 5. Separate analyses might also be carried out for urban and nonurban areas of the country. Unfortunately, socioeconomic and demographic data comparable to those available for counties cannot be compiled for individual urban areas, as distinct from the counties of which they are a part. Urbanization, however, may be conceived as a process that subjects populations to certain social processes. From this perspective, "urbanicity" can be viewed not just as the concentration of population but as the concentration of both the working-age and more highly educated populations, in particular; as greater exposure to newspapers and (to a lesser extent) television; as greater access to social services, including medical care; and, finally, as increased intergroup contact. These social processes are reflected in the strong 552 Nationality in Yugoslavia correlations displayed in Appendix B. They may be said to be the politically most salient dimensions of urbanization, while the process of establishing the physical infrastructure characteristic of urban places is simply a reflection of the level of material development. The proportion of apartments in each county linked to a public sewerage system in 1981, for example, may be considered a strong indicator of the presence of such infrastructure. It correlated much more strongly with the overall level of development of the commune (as measured by per capita income in 1980) than with its 1981 population density. Even when density was held constant, this relationship remained strong. Moreover, it held in every region, both developed and underdeveloped. Thus, rather than grouping the cases into urban and nonurban counties, we accounted for the effects of urbanization by focusing on its politically salient component processes, as measured by the relevant explanatory variables. 6. The multiple linear regression model for a dichotomous (0,1) dependent variable is termed the linear probability model, and this model is often compared unfavorably with other techniques. Criticisms of it concern (1) the lack of efficiency (minimum variance) of its coefficient estimates, which is obviated by shifting from ordinary least squares to weight least squares estimation; and (2) the possibility of out-of-range (i.e., outside [0,1]) linear predictions, which is obviated by employing a nonlinear function that is bounded between 0 and 1 (e.g., logistic). However, if the data are aggregated so that the dependent variable becomes the proportion of cases in each group exhibiting the behavior of interest (i.e., the proportion who declare Yugoslav identity in each county), "so long as the sample sizes in each cell are sufficiently large, and provided p{ is not equal to 0 or 1, the linear probability model is not subject to the criticisms made earlier (Maddala 1983, 29; emphasis added). The complete census count for a Yugoslav country is certainly a large enough sample size, and our predicted proportions using the linear probability model are within range. Our coefficient estimates are not fully efficient owing to mild heteroscedasticity, but that hardly matters in the context of a "sample" consisting of the entire population of counties, in which virtually the entire population of individuals was contacted. Thus, we present results throughout this paper based on the linear probability model for aggregate data estimated via ordinary least squares estimation. 7. Another approach to determining the cross-time validity of the findings might be to develop a dynamic model, using changes over time in the values of our explanatory variables to explain changes in the proportion of "Yugoslavs." However, while there has been substantial change over the period 1971-81 (for example, in the level of material and social development of the counties and in the proportion of the population with higher educa- tion), very strong rank order correlations between 1971 and 1981 values of our predictor variables suggest that the relative positions of the counties on these indicators have remained substantially unchanged. These two approaches, therefore, yielded essentially the same results with respect to identifying important explanatory variables. 8. Data not shown. Regression tables available on request. 9. Jug Grizelj, "Koliko ste zadovoljni radom SIV" [How satisfied you are with the work of the FEC], NIN, 20 April 1986; Ljuba Stojic, "Na granici odricanja" [At the limit of self-denial], NIN, 10 October 1987. 10. Marinko Culic, "Peticionasi brisu ustav" [Petitioners erase the constitution], Danas, 29 March 1988; idem, "Kakav ustav hoce pisci?" [What kind of constitution do the writers want?], Danas, 3 May 1988. References Bakic, Ibrahim. 1985. Jugosloventsvo od ide ja do ostvarenja [Yugoslavism from idea to realization]. Belgrade: Isro Privredno Finansijski Vodic. Bertsch, Gary K. 1972. "A Cross-national Analysis of the Community-building Process in Yugoslavia." Comparative Political Studies 4 (4): 438-60. Burg, Steven L. 1983a. Conflict and Cohesion in Socialist Yugoslavia. Princeton: Princeton University Press. Burg, Steven L. 1983b. The Political Integration of Yugoslavia's Muslims: Determinants of Success and Failure. Carl Beck Papers in Russian and East European Studies no. 203. Pittsburgh: University of Pittsburgh. Burg, Steven L. 1986. "Elite Conflict in Post-Tito Yugoslavia." Soviet Studies 38 (2): 170-93. Burks, Richard V. 1971. The National Problem and the Future of Yugoslavia. Rand Report no. P-4761. Santa Monica, CA: Rand Corporation. Cohen, Lenard, and Paul Warwick. 1983. Political Cohesion in a Fragile Mosaic: The Yugoslav Experience. Boulder, CO: Westview. Converse, Phillip E., and Roy Pierce. 1986. Political Representation in France. Cambridge: Harvard University Press. Cvjeticanin, Vladimir, ed. 1984. Klasno-socijalna struktura saveza Komunista jugoslavije [The Socio-class structure of the League of Communists of Yugoslavia]. Belgrade: Komunist. Deutsch, Karl W. 1953. Nationalism and Social Communication. Cambridge: Technology Press of MIT. Deutsch, Karl W. 1957. Political Community and the North Atlantic Area. Princeton: Princeton University Press. Deutsch, Karl W. 1961. "Social Mobilization and 553 Nationality in Yugoslavia Political Development." American Political Science Review 55:493-507. Dugandzija, Nikola. 1984. Jugoslavenstvo [Yugo-slavism]. Belgrade: Mladost. Easton, David. 1975. A Re-assessment of the Concept of Political Support. British Journal of Political Science 5:435-57. Etzioni, Amitai. 1965. Political Unification. New York: Holt, Rinehart & Winston. Here, Sergej. 1987. "Ethnic Attitudes of Youth in Yugoslavia: The Question of Ethnic Yugoslav-ism." Novi Sad University, Yugoslavia. Typescript. Hanushek, Eric A., John E. Jackson, and John F. Kain. 1974. "Model Specification, Use of Aggregate Data, and the Ecological Correlation Fallacy." Political Methodology 1:89-107. Hechter, Michael. 1975. Internal Colonialism: The Celtic Fringe in British National Development, 1536-1966. Berkeley: University of California Press. Janowitz, Morris. 1983. The Reconstruction of Patriotism: Education for Civic Consciousness, Chicago: University of Chicago Press. Kornberg, Allan, Harold D. Clarke, and Laurence LeDuc. 1978. "Some Correlates of Regime Support in Canada." British Journal of Political Science 8:199-216. Kornberg, Allan, Harold D. Clarke, and Marianne C. Stewart. 1980. "Public Support for Community and Regime in the Regions of Contemporary Canada." American Review of Canadian Studies 10:75-93. Lichtman, Allan J. 1974. "Correlation, Regression, and the Ecological Fallacy." Journal of Interdisciplinary History 4 (3): 417-33. Lijphart, Arend. 1975. The Politics of Accommodation. 2d, rev. ed. Berkeley: University of California Press. Lijphart, Arend. 1977. Democracy in Plural Societies. New Haven: Yale University Press. McRae, Kenneth, ed. 1974. Consociational Democracy. Toronto: McClelland & Stewart. Maddala, G. S. 1983. Limited-Dependent and Qualitative Variables in Econometrics, New York: Cambridge University Press, Steven L. Burg is Associate Professor of Politics and Michael L. Berbaum is Assistant Professor of Psychology, Brandeis University, Waltham, MA 02254, 554 Matvejevic, Pedrag. 1983. Jugoslavenstvo donas [Yugoslavism today]. Zagreb: Globus. Miric, Jovan. 1984. Sistem i kriza [System and crisis]. Zagreb: Cekade. Müller, Edward N., and Thomas O. Jukam. 1977. "On the Meaning of Political Support." American Political Science Review 71:1561-95. Müller, Edward N. 1979. Aggressive Political Participation. Princeton: Princeton University Press. Müller, Edward N., Thomas O. Jukam, and Mitchell A. Seligson. 1982. "Diffuse Political Support and Antisystem Political Behavior: A Comparative Analysis." American Journal of Political Science 26:240-64. Müller, Edward N., and Carol J. Williams. 1980. "Dynamics of Political Support-Alienation," Comparative Political Studies 13:33-59. Obradovic, Vladimir. 1984. "Jugoslovenstvo u stavovima i izjasnjavanju omladine" [Yugoslavism in the beliefs and declaration of youth], in Mladi i nacionalizam [Young people and nationalism], ed. Dusan Davidovic. Belgrade: Mladost. Rae, Douglas W., and Michael Taylor. 1970. The Analysis of Political Cleavages. New Haven; Yale University Press. Raic, Aleksandar. 1982. "Jugosloveni u Vojvodini" [Yugoslavs in Vojvodina]. Nase Terne 26 (10): 1713-22. Rusinow, Dennison. 1977. The Yugoslav Experiment 1948-1974. Berkeley and Los Angeles: University of California Press. Savezni Závod za Statistiku [Federal Institute for Statistics]. 1981. Strucno-metodolosko uputstvo za organizovanie i sprovodjenje popisa [Technical-methodological handbook for organization and implementation of the census]. Rev. ed. Belgrade: Federal Institute for Statistics. Schelling, Thomas C. 1978. Micromotives and Macrobehavior. New York: W. W. Norton. Shoup, Paul. 1968. Yugoslav Communism and the National Question. New York: Columbia University Press. Vuskovic, Boris. 1982. Tko su Jugoslaveni?" [Who are the Yugoslavs?], Nase Terne 26 (10); 1702-12.