Nationalities Papers, Vol. 32, No. 1, March 2004 One Ukraine or Many? Regionalism in Ukraine and Its Political Consequences Lowell W. Barrington & Erik S. Herron Intra-state regional differences are a central topic in the study of European and Eurasian politics. In Ukraine, regional differences have proven to be powerful predictors of mass attitudes and political behavior. But what does the “regional factor” in Ukrainian politics represent? Is it simply the result of compositional effects, or are the regional differences more than just a sum of other demographic factors correlated with geographic divisions? When analyzing regional divisions as an explanatory variable, what are the implications of employing different regional frameworks? In this article, we demonstrate how geographic divisions in the country hold up even when others factors—such as ethnicity and language use—are con- trolled for. As part of this inquiry, we compare the results of three competing regional frameworks for Ukraine: one with two regions, one with four regions and onewitheightregions.Whiletheeight-regionframeworkisuncommoninstudiesof Ukraine, the decision to examine eight regions is supported by historical, economic and demographic arguments, as well as by the results of the statistical analyses presentedinthisarticle.ScholarswhohavefocusedonfewerregionsinUkrainemay haveunderestimatedtheeffectsofregionaldifferencesandmissedinterestingstories about intra-state variation in Ukrainian attitudes and voting behavior. The results of this study carry important implications not only for the study of Ukraine but also for those interested in intra-state regional divisions across Europe and Eurasia. Itistakenasatruismbymostscholarsofpost-CommunistEuropethatinUkraine, more than in most countries, geography matters. The existence of powerful intra- state regional divisions in mass attitudes and political behavior (e.g. voting) in Ukraine has been a theme of numerous studies (Birch, 2000; Kubicek, 2000; Craumer and Clem, 1999; Barrington, 1997; Holdar, 1995). But what does it mean that there are strong regional differences in attitudes and behavior? Are the regional differencessimply“compositionaleffects”(capturingotherthingsthatscholarsclaim are important in Ukraine such as language and ethnicity), or are they capturing underlyingdivisionsthatcanbeadequatelydisentangledfromethnicityandlanguage andexertanindependenteffect?Thisquestionisafocusofourarticle.Butweargue that it is not enough to demonstrate that regional divisions exist even when language and ethnicity are controlled for, as others have shown in Ukraine (Barrington, 2002; 2001; 1997; Birch, 2000; Wilson and Birch, 1999). Rather, the central question of ISSN0090-5992print;ISSN1465-3923online/04/010053-342004AssociationfortheStudyofNationalities DOI:10.1080/0090599042000186179 L. W. BARRINGTON & E. S. HERRON our article is how different regional frameworks affect analyses of Ukrainian mass attitudes and political behavior. Although one might argue that the decision about how many regions to examine andwhichoblaststoincludeinwhichregionissubjective(asevidencedbythelarge number of different regional arrangements examined by scholars of Ukraine), we demonstrate in this article that such a decision has consequences. As a result, scholars must carefully decide how best to delineate regions in Ukraine based on historical, economic and demographic features. Scholars must also take into account the concept of region as an intermediate geographic entity: regions must be large enough to justify the label “region” but not so large as to combine together significantly different areas in terms of history, economics and demographics; very small regions (with the region in question created from a single oblast, “autonomous region” or single city) and very large regions (half the country) must involve compelling cases for their treatment as “regions.” We divide the article into three sections. In the first section, we review the question of regionalism in Ukraine, specifically addressing a recent challenge to the importance of this cleavage. Next, we address political attitudes, with one section presentingresultsoftheeffectsofviewsabouttheethnic“other”andthesubsequent section presenting findings about support for the government and political system. Lastly, we assess the role of region in parliamentary and presidential elections. Regionalism in Ukraine There is a debate over the way that geography intersects with political attitudes and behavior in Ukraine. Most scholars suggest that reasonably well-defined geographic divides exist. Others suggest that there is substantial intra- and inter-regional variation that we must account for. Indeed, we would expect that there is some bleeding along borders; the “regional factor” and its influence on political attitudes and behavior may not coincide perfectly with existing intra-Ukraine territorial borders. Yet, we also have reasons to believe that “region” may capture important differences among Ukrainians. Do Regions Matter in Ukraine? Few scholars of Ukraine would argue that regional divisions in the country are unimportant. But, O’Loughlin (2001) has made a case that the emphasis on regional divisions in the country is flawed. While O’Loughlin presents several arguments in his article questioning whether regional effects are “bogus,” two claims in his critique are most damaging to the enterprise of studying regional effects in Ukraine: (1)apparentregionaldifferencesmaskimportantvariationwithinregions,and(2)the division of the country into regions creates artificially sharp internal boundaries that miss the “blurring” of regional divisions in Ukraine.1 54 ONE UKRAINE OR MANY? His first point depends on the idea that any apparent regional effects are under- mined if analysis of localities and other lower levels of geographic scale indicates significant variation within the larger geographic unit. As O’Loughlin (2001, p.5) puts it, “Local and oblast-level patterns could show different trends and support contradictory hypotheses regarding the significance of the geographic effect in Ukraine.” While the question of scale is a crucial one in geography (and a focus of our comparison of different regional approaches in this study), the idea that the existence of intra-regional variation weakens the argument for the existence of regional divisions is unconvincing. Within any group of respondents (separated from other respondents on the basis of regional, linguistic, gender or other differ- ences), there will be variation.2 O’Loughlin’s point is only valid if there are compelling theoretical, historical and empirical reasons to consider very small geographicareasasthefocusofthegeographicalanalysisattheexpenseofthelarger regions that scholars have traditionally examined. He offers no such compelling case. In addition, if there is indeed too much internal variation and too little external variation among the regions, one would not expect to see a statistically significant regional effect (whether the intra-region variation was systematic or simply noise, too much of it would lead to a lack of statistical significance on regression coefficients). Numerous findings of statistically significant regional effects in the existing literature on Ukraine indicates, to the contrary, that something is going on at the regional level, even taking into account internal variation. While it is, therefore, important to keep in mind that not everyone in a certain region will think the same (just as not every member of a particular ethnic group or linguistic group will see eye to eye), this alone does not justify ignoring strong evidence of regional differences in Ukraine. O’Loughlin’s second argument is more compelling but, we contend, also ultimately flawed. It is quite reasonable to expect that the existing political- administrative borders upon which definitions of Ukrainian regions are based do not adequately delineate differences between regions. Scholars base their regional divi- sions on these boundaries, but the divisions researchers employ may not be organic and may inappropriately divide areas that have similar features that contribute to relativelycohesivepoliticalattitudesandbehavior.Further,wewouldexpectaripple effect along natural regional borders and across borders with neighboring countries. Attitudinal or behavioral effects that emanate from one region may be difficult to distinguish from other regions as populations become more proximate to one another. At the same time, analyses that examine geographic factors require lines to be drawn somewhere. After all, if we are to hold that attitudinal blurring around regional boundaries invalidates the use of region as an explanatory variable, multi- national studies of survey data that control for the country of residence of the respondents must likewise be discarded. The blurring of attitudes between age 55 L. W. BARRINGTON & E. S. HERRON groups is also likely (with 59-year-olds having more in common with 60-year-olds than with 50-year-olds). Yet we do not dispense with the concept of age cohorts because of such attitudinal blurring. Thus, although we acknowledge the concerns about demarcating regional boundaries, we do not believe that such concerns outweigh the benefits of examining the extent to which regional differences in attitudes and behavior exist in Ukraine. Strong Regional Differences or Compositional Effects? Assumingoneissatisfiedthatexaminingtheattitudinalandbehavioralconsequences of regional differences is a worthy endeavor, the next step is to consider two contrasting ideas of what any apparent regional effects on political attitudes and behaviorrepresent.Eithertheyarecompositionaleffectsortheyarerepresentativeof anunderlyingsetofcohesiveattitudesandbehaviorsthatmayinfactreflectdiffering regional political cultures. Testing for which of the two ideas of regional differences is more valid is rather easy, through analysis that controls for the various things that people would assume to be the compositional components. O’Loughlin (2001, p.8) points out that in Ukraine these “compositional groups” include classes, religions, ethnicgroupsandurbanpopulations,thoughonewouldcertainlyalsowanttocontrol for language use as well as include other traditional demographic control variables such as gender and age in any analysis. While O’Loughlin (2001, p.8) claims that “the unresolved question is whether the east–west divide remains visible when these factors are taken into account,” studies previous to his (e.g. Birch, 2000; Barrington, 1997) have demonstrated the existence of regional effects even when controlling for possible compositional components. This article presents results that support the existence of strong regional divisions separate from any compositional effect. Demonstrating that regional differences appear even when testing for composi- tional effects does not answer the question of why this is the case. We make two claims. First, the regional frameworks we examine can be justified in ways that are not easily captured by individual-level compositional variables (e.g. differing histor- ies of foreign rule). Second, an extensive literature exists that highlights the importance of geographically based contextual factors in shaping attitudes and behavior (see, for example, Huckfeldt and Sprague, 1993; Huckfeldt, 1974). This literature argues that living in a particular setting shapes attitudes over time. People who move from one region of the United States to another, for example, may see their attitudes change in line with the attitudes of those now around them. As Ulbig (1999, p.1) puts it, “People learn from and adjust to the political and social contexts surrounding them.” This means that even if compositional effects were initially behind the emergence of regional differences, regional attitudes can take on a life of their own over time, separate from their compositional foundation. 56 ONE UKRAINE OR MANY? How Many Regions? If controlling for possible composition effects reveals geographic differences, the question still remains: how many regions to include in analyses of attitudinal variation and political behavior? As we demonstrate in this article, different regional frameworks can produce different substantive findings. This is true not only about the effects of regional divides but also about the effects of other variables, once region is included in the investigation. In the statistical analyses below, we consider three competing regional breakdowns. The first is a two-region option, with the oblasts of Ukraine divided—as much as possible—between east and west by the Dniepr river. Such an approach has been used by many scholars, especially those focusing on electoral results.3 The second regional arrangement we examine, even morepopularwithscholars,isafour-regionvariant—madeupofwest,central,south and east regions. This designation of oblasts into larger regions is based on Arel (1992). Finally, we consider an eight-region option for Ukraine given the country’s economic development patterns, divergent historical experiences, and demographic features. Such characteristics of Ukraine support the idea that its oblasts cannot easily be combined into only two or even four regions.4 Becausetheeight-regionframeworkislesscommoninstudiesofUkrainethanthe two- or four-region approaches,5 we will briefly present an argument for this arrangement prior to examining the consequences of choosing one framework over another. The basic justification for an eight-region approach is that two and four regions inadequately differentiate among areas with different historical, economic anddemographicfeatures.Whiletheidentificationofeightregionsmaynotperfectly divide Ukraine into areas with cohesive social and political identities, this approach is more likely to meaningfully cluster together geographic units. The four-region framework (see Arel, 1992) places Donetsk, Luhansk, Kharkiv, Dnipropetrovsk and Zaporizhzhia in the east. We acknowledge that these oblasts are distinct from the rest of the country, but we also feel there are solid reasons to consider these oblasts as comprising two separate regions. In the eight-region approach examined in this study, therefore, two of these oblasts—Donetsk and Luhansk—form the “east” region.6 They both border the Russian Federation. This areatendstobemoreindustrial,urban7andRussifiedthanotherareasofthecountry, but also more Russified than Kharkiv, Dnipropetrovsk and Zaporizhzhia (Hesli, 1995). According to Arel (1992, 13), linguistic Russification has combined with the high percentage of ethnic Russians to create “solid majorities of Russophones” in Donetsk and Luhansk (68% and 64%, respectively).8 Kharkiv, Dnipropetrovsk and Zaporizhzhia provinces constitute the “east-central” region in the eight-region analyses.9 Although they are heavily industrial and Russified, they are less demographically Russian than their neighbors to the east. Dnipropetrovsk is particularly worthy of note in this regard, since it is the second most industrialized oblast in Ukraine10 but is the least Russified of the five oblasts 57 L. W. BARRINGTON & E. S. HERRON in the east and east-central regions. In addition, two of the three east-central oblasts—Dnipropetrovsk and Zaporizhzhia—do not border Russia. Kharkiv does, although it is much less Russified than the oblasts to its east (Hesli, 1995). The “north-central” region of the eight-region framework is an intriguing part of the country. Poltava, Kirovohrad, Cherkasy, Kyiv, Chernihiv and Sumy oblasts, along with the city of Kyiv, are included in this region.11 The area did not come underRussiancontroluntilthemiddle1600stothelate1700s(Szporluk,1997).The area is more industrial than much of the west but less so than the east (Hesli, 1995). WiththeexceptionofthecityofKyiv,theareaalsohaslowerpopulationdensityand islessethnicallyandlinguisticallyRussianthantheareastoitseast(seeArel,1995). Although Chernihiv and Sumy border Russia, they are much less urban, industrial andRussifiedthantheeastoreast-centralregions,andtheysharemanydemographic and historical features with the oblasts to their south. The “southern” region of Ukraine in the eight-region analysis is composed of Kherson, Odesa and Mikolaiv provinces.12 This area does not abut Russia (the RepublicofKrymistreatedasitsownregion;seebelow).Also,theRussiansdidnot gain control of it until the late 1700s; much of it had been under Ottoman Turkish control (Szporluk, 1997; Arel, 1992). The area is also less urban, industrial and ethnically Russian than the east (see Hesli, 1995).13 Theeight-regionframeworktreatsKrym(alongwithitscityofSevastopol,oneof the two constitutionally designated cities of special significance in Ukraine) as a distinctregion,whichisuncommoninotherscholarship.14Therearegoodreasonsto treat it as a separate region, especially as one considers whether the geographic differences in the country are capturing underlying regional cultures. Krym was the only part of Ukraine that did not strongly support independence in the 1991 referendum.15 Russians comprise the majority of the region’s population, nearly half of the Ukrainians there speak Russian as their first language (Solchanyk, 1994), and it was last area to formally join Ukraine (by act of the Soviet government, it was transferred from the Russian Soviet Federated Socialist Republic [RSFSR] in 1954). By any standard, it is the least Ukrainian part of the country (see Barrington, 2002; Shaw, 1994). The provinces of Zhytomyr, Vinnytsia, Khmelnytskyi, Rivne and Volyn form the “west-central” region.16 These oblasts were separated from both the central and western regions in the eight-region framework in part to gauge the extent to which thereisacertaindegreeofregional“blurring”(Barrington,2002)asonemovesfrom thecenterofthecountrytothefarwest.Buttherearealsosolidhistorical,economic and demographic reasons for treating them as a distinct region. Unlike the oblasts to their south and west, these five oblasts fell under control of Russia as a result of the partitioning of Poland. Russia gained the territory that includes Zhytomyr, Vinnytsia andKhmelnytskyiduringthesecondpartitionin1793,andRivneandVolynwiththe thirdpartitionin1795.17Inaddition,theoblastsofthisregionallhavebelowaverage population densities, ones below those of the oblasts to their south and west. 58 ONE UKRAINE OR MANY? In the eight-region framework, we also separate Chernivtsi and Zakarpatia— labeling it the “southwest”18—from the rest of the west region. These oblasts have lowlevelsofindustrialization,althoughinterestinglyrankatthebottomofthelistof oblasts in terms of agricultural production as well (Hesli, 1995). They are both, however, above the average in oblast population density. More important, they both share long borders with neighboring countries (Slovakia, Hungary and Romania in the case of Zakarpatia, and Romania and Moldova in the case of Chernivtsi). Partly as a result of their location, and partly for historical reasons19 the level of “Ukrainianness”intheseoblastsismuchlowerthanthatoftheoblaststotheirnorth: Lviv, Ternopil and Ivano-Frankivsk. These remaining three oblasts make up the “west” region.20 These three are often what scholars think of when they discuss the west of Ukraine. This region, what some call “Galicia” to this day, was part of the Polish territory of the Austro- Hungarian Empire (coming under the control of Austria as a result of the first partition of Poland in 1772) and controlled by independent Poland in the interwar period (Arel, 1992). The region is average to above average in population density. It is more ethnically Ukrainian than the oblasts to its south, and it was the heart of anti-RussianUkrainiannationalismintheyearsprecedingandfollowingthecollapse of the Soviet Union. Thus,wearguethattherearestronghistorical,economicanddemographicreasons to consider more regions than is the norm in studies of Ukraine. In Table 1, we present the three alternative regional arrangements that we examine statistically in the next several sections, with the oblasts (and Krym) and the cities of Kyiv and Sevastopol (which are given special status in the Ukrainian Constitution) mapped into their respective regions. Mass Attitudes: Regional Differences in Views about the “Ethnic Other,” and about the Ukrainian Government and Political System In the following sections, we analyze survey data about views of the “ethnic other” in Ukraine, support for the political system and support for the government. The survey data analyzed in this paper come from a late 1998 nationwide survey conducted in Ukraine.21 The statistical analysis of these data support the notion that regional effects are not simply compositional effects, as well as highlighting the implications of employing differing regional frameworks. Views of the Ethnic Other: Factors Affecting Individual-Level Variation Ukraine is a country in which ethnic tensions were thought to be a serious potential problem, yet where serious ethnic conflict did not develop. It is, therefore, a useful case for those interested in ethnic conflict. While scholars remain concerned about inter-ethnic relations in Ukraine, particularly between ethnic Ukrainians and the 59 L. W. BARRINGTON & E. S. HERRON TABLE 1 Distribution of the oblasts, cities of special significance and the Krym Republic into the three regional arrangements examined in this paper Republic/oblast/city 2 divisions 4 divisions 8 divisions Donetsk East East East Luhansk East East East Kharkiv East East Eastcentral Zaporizhzhia East East East-central Dnipropetrovsk East East East-central Krym (Crimea) East South Krym Sevastopol East South Krym Kherson East South South Mikolaiv West South South Odesa West South South Chernihiv East Central North-central Sumy East Central North-central Poltava East Central North-central Cherkasy West Central North-central Kirovohrad West Central North-central Kyivska Oblast West Central North-central Kyiv City West Central North-central Khmelnytskyi West Central West-central Zhytomyr West Central West-central Vinnytsia West Central West-central Rivne West West West-central Volyn West West West-central Ivano-Frankivsk West West West Lviv West West West Ternopil West West West Chernivtsi West West Southwest Zakarpatia West West Southwest country’s large ethnic Russian population, a theme of some studies of Ukraine has beenthe“dogsthatdidn’tbark”phenomenon—whydidnosignificantethnicconflict emerge in Ukraine (see, for example, D’Anieri, 1998; Dawson, 1997)? While these studieshaveconsideredavarietyofexplanations,theyhavenotexaminedthedegree of, and causal factors shaping, ethnic stereotypes in Ukraine. As a result, our understanding of ethnic relations in Ukraine is incomplete. In addition, the lack of such a study means that Ukraine has not been utilized in the way it can be—as a valuable case for developing a broader understanding of ethnic stereotyping in multi-ethnic states. Aggregate level findings from the analysis of the survey data indicate that views of the ethnic other in Ukraine are generally positive, with Russians viewing Ukrainians a little more positively than Ukrainians view them. Yet, statistics such as 60 ONE UKRAINE OR MANY? the mean score on a scale of stereotyping (see Barrington, 2001b) can tell us little about what shapes such attitudes at the individual level. Given that there is a noticeablevariationinviewsoftheethnicotheramongtherespondents,wefocuson trying to understand the factors that influence such variation.22 Regional divides There is a sense among scholars of stereotyping that how often in-groups and out-groups interact must be taken into account. Scholars disagree aboutwhether“proximity”totheethnicotherwillfostertoleranceorresentment(see Burns and Gimpel, 2000; McAndrew et al., 2000), but nearly all consider it to be an important factor.23 One could think of several ways to capture the “geographic proximity”conceptdependingontheleveloneismostinterestedin,but—forreasons discussed earlier in the paper—regions make a great deal of sense in the case of Ukraine(seealsoBremmer[1994]onethnicstereotypesinUkrainebyregions).This is a point that was made strongly by participants in two focus groups conducted in 1999 in Ukraine as well (see Barrington and Herron, 2002). Existing studies such as Burns and Gimpel (2000) indicate that “border” areas may be more likely to hold negative views of minorities than other regions. Before examining the effects of region through a comparison of different regional coding schemes in a multivariate estimation, it is necessary to consider the role of other basic, “demographic”24 factors in shaping stereotypes in Ukraine. Many of these features are both common to statistical models explaining individual-level attitudinal variation and are discussed by scholars specifically interested in stereo- typical views of the “other.” Much of the justification for the inclusion of these variables echoes the work of Burns and Gimpel (2000), who estimate models of racial stereotyping in the United States in the 1990s. Ethnicity Another logical factor to examine when considering demographic deter- minants of ethnic stereotyping is ethnicity itself. In a multi-ethnic state, one group maynotbenearlyashostiletowardasecondgroupasthatsecondgroupistoit.This is especially true in cases of a majority and large minority. In addition, a majority group may be the main ethnic other for a variety of ethnic groups, many of which may hold very different views about that majority. In Ukraine, the mean scores on a stereotyping scale constructed from the survey data and comments from the focus groups (see Barrington and Herron, 2002) indicate that Russians and Ukrainians do not see each other in equally positive lights. But while the groups differed, they did not differ as much as one might have expected. In addition, such variation at the group level may not hold up at the individual level once other factors are controlled for. In Ukraine, ethnicity has not been “triggered” by elites to the extent that it has in other post-Soviet states (Barrington, 2002). Part of the reason for this is that ethnicity is not as simple to trigger in settings where the ethnic lines are blurred. Due to intermarriage (Rapawy, 1997; Pirie, 1996) and language use (see below), Ukraine could be seen as a case of 61 L. W. BARRINGTON & E. S. HERRON ethnic blurring (see also Smith and Wilson, 1997). As Ponarin (2000, p.1535) puts it, “Even though the boundaries between Western Ukrainians and Russians may seemasstrongasthosebetweenRussiansandEstonians,theboundariesbetweenthe various Ukrainian groups in actual contact are quite fluid.” Analysis of survey data inBremmer(1994)pointstothisblurring,sinceconspicuousnumbersofrespondents in Simferopol and, to a lesser extent, Kyiv were unable to identify the ethnic make-up of their neighborhoods. This was not the case in Lviv. Language One reason for the perception of a “fluid” or “blurred” nature of ethnicity in Ukraine is that linguistic and ethnic lines are not complementary. Specifically, the existence of large numbers of Russian-speaking Ukrainians in the country cuts into the ethnic divisions.25 This has led scholars to claim that language is, in fact, the main “fault line” in Ukraine (Arel, 1998) and that Russian-speaking Ukrainians may think more like Russians than like their Ukrainian-speaking co- ethnics(CraumerandClem,1999).Supportersofthisviewfeelthatifonespeaksthe same language as the ethnic other, it is likely that one would view members of that ethnic group more favorably (see also Poppe and Hagendoorn, 2001). But while the language fault line idea makes sense, several studies call into question its importance. Bremmer (1994) points out that even in areas where Ukrainians are not generally fluent in Ukrainian, they often claim to be, showing a sense of attachment to their “own” culture and a desire to express this attachment. In addition, the language variable has not always held up in statistical analyses of Ukrainian mass attitudes which include both ethnic and regional controls (see Barrington, 2002; 2001a; 1997). Urban/rural divisions The percentage of Russians also varies greatly based on size oflocalityinUkraine.ThelargeoblastcapitalcitieshavesignificantlymoreRussians than smaller towns around them, and these differences have been associated with individual-level differences in post-Soviet political attitudes (Craumer and Clem, 1999;Melvin,1995).ThisisconsistentwiththepreliminarydiscussioninBurnsand Gimpel (2000), as well as the results of their study which indicate that urban residents, other things constant, are more favorable in their views of other ethnic groups. Thus, one might also expect to find variation in stereotypes as the result of differences in size of locality in Ukraine, and, as a result, locality size is also included in the model estimated below. Religiousbeliefs Likelanguage,religionisavariablethatcross-cutsethnicidentity in Ukraine. Not only are there differences within ethnic groups between believers and non-believers, but believers among ethnic Ukrainians differ in terms of denom- ination (Gee, 1995). Especially in the far west of the country, there is a large Ukrainian Uniate population, while most Ukrainian believers in the rest of the country would associate themselves with the Ukrainian Orthodox or Russian 62