Determinants of Regional Political Distinctiveness

Nationalism and Ethnic Politics

ISSN: 1353-7113 (Print) 1557-2986 (Online) Journal homepage: http://www.tandfonline.com/loi/fnep20

Determinants of Regional Political Distinctiveness Britt Cartrite & Dan Miodownik To cite this article: Britt Cartrite & Dan Miodownik (2016) Determinants of Regional Political Distinctiveness, Nationalism and Ethnic Politics, 22:2, 119-148, DOI: 10.1080/13537113.2016.1169057 To link to this article: http://dx.doi.org/10.1080/13537113.2016.1169057

Published online: 04 May 2016.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=fnep20 Download by: [185.120.125.129]

Date: 04 May 2016, At: 19:40

Nationalism and Ethnic Politics, 22:119–148, 2016 Copyright © Taylor & Francis Group, LLC ISSN: 1353-7113 print / 1557-2986 online DOI: 10.1080/13537113.2016.1169057

Determinants of Regional Political Distinctiveness BRITT CARTRITE Alma College

DAN MIODOWNIK

Downloaded by [185.120.125.129] at 19:40 04 May 2016

Hebrew University

Scholarly research exploring the phenomenon of regional distinctiveness in Europe, since at least the 1960s, has generated a variety of competing theories to explain the phenomenon, including the following: the persistence of linguistic distinctiveness; the impact of economic distinctiveness; and remoteness. Often these studies operationalize “regional distinctiveness” in different ways, impeding the evaluation of different types of theories against one another. This study develops a novel measure for regional distinctiveness, applied to 161 regions in 11 European countries from 1990–2014, and demonstrates that language, economics, and remoteness work through regional parties to generate regional political distinctiveness, while only linguistic distinctiveness also has a direct effect on such distinctiveness.

INTRODUCTION On 18 September 2014, 84.6% of Scottish voters, a record for elections in the United Kingdom since the establishment of universal suffrage, turned out to respond to a single question: “Should Scotland be an independent country?” While the majority “no” vote was not surprising, that 44.7% of voters responded “yes” suggests that, despite having been part of the United Kingdom since 1707, politics in Scotland remain markedly distinct from those of England. The following day the Catalan parliament voted to hold a similar referendum on the status of Catalonia within Spain, scheduled for

Address correspondence to Britt Cartrite, Alma College, Swanson Academic Center 353, 614 W. Superior St., Alma, MI, 48801, USA. E-mail: [email protected] Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/fnep. 119

Downloaded by [185.120.125.129] at 19:40 04 May 2016

120

B. Cartrite & D. Miodownik

7 November 2014. Although the formal vote was deemed unconstitutional and therefore forbidden by the Spanish government, an informal referendum was nonetheless held with, perhaps not surprisingly, some 80% of those participating voting for independence. Scotland and Catalonia represent only two of a large number of regions in which voters regularly behave in ways that are often distinct from the nation as a whole, in their willingness to participate, their support for regional parties, or even by simply voting for national parties in distinctive patterns. The persistence, and indeed growth, of such distinctiveness despite centuries of state- and nation-building projects would surprise many scholars and policy makers of the 19th and early 20th centuries, from Jacobins to Allied leaders following World War I, who argued that the homogenization of the polity was both desirable and inevitable. Yet, by the 1960s, it had become clear that, despite long periods of democracy, capitalist economic development, and advances in communications and transportation technology, regional political distinctiveness persisted in many European states, leading scholars such as Stein Rokkan, Derek Urwin, Ernest Gellner, and others to try and account for the persistence of local distinctiveness in the late 20th century. The resultant and ongoing body of scholarship focuses on a variety of explanatory dynamics explaining regional distinctiveness, in particular the impact of cultural and economic distinctiveness and remoteness from the core. However, there is no agreed-upon metric by which regions are evaluated in terms of political distinctiveness, nor indeed any overall agreement on how “regions” are to be conceptualized. As a result, studies often differ on the number and nature of regions studied, measures of “distinctiveness” deployed, and perhaps not surprisingly come to widely varying conclusions regarding regional distinctiveness. This study attempts to synthesize some of the major themes in the regionalism literature with an eye towards developing a more comprehensive analysis of regionalism that will eventually allow for the study of all regions within Europe, taking into account cultural, economic, and geospatial factors. We begin with an exploration of the relevant literatures that generate operationalizations and expectations regarding the persistence of political distinctiveness. In the “Research Design” section, we next describe the development of a new index of regional political distinctiveness that embeds voter turnout as part of the measure, allowing for a broader understanding of “political distinctiveness” than measures that simply focus on variation in votes for particular parties, which we subsequently apply to 161 regions across 11 European Union member states. In addition, we explore regional-level measures for linguistic distinctiveness, economic distinctiveness, remoteness, and the relative success of regional political parties. In the “Analysis” section, we apply a “path analysis” statistical model,

Determinants of Regional Political Distinctiveness

121

demonstrating that, while language, economics, and distance each are mediated by regional parties in driving regional political distinctiveness, only linguistic distinctiveness has a significant independent effect on regional political distinctiveness; the resultant model represents an important advance in studies of regionalism, in that it establishes the relationships between factors that are rarely considered together. The final section concludes the study.

Downloaded by [185.120.125.129] at 19:40 04 May 2016

THE PERSISTENCE OF REGIONAL POLITICAL DISTINCTIVENESS In a recent review of the party systems nationalization literature, Arian Schakel notes on the vigorous debate regarding the dynamics of voting behavior over space and time, with studies focusing on differential changes in support for individual parties over time (“dynamic nationalism”), variation across space in party success (“distributional nationalism”), and across institutional levels (“multilevel nationalism”).1 This literature, which derives from the expectation that the formation of the modern nation-state necessarily entails a process of state-nation building, argues that modern states centralize and consolidate power by increasingly if unevenly homogenizing subnational regional political behavior as national politics and pressures come to dominate local political discourses. While research in this area dates back to at least Elmer Schattschneider,2 Daniele Caramani’s3 seminal work on party system nationalization in Europe provides an excellent summary and exploration of this field. Building from the works of Albert O. Hirschman4 on voice and exit and Stein Rokkan and Derek Urwin’s5 research on territoriality and regionalization, Caramani argues that the consolidation of the modern state entails the dual processes of penetration and standardization of peripheries and the “vertical dislocation of issues, organizations, allegiances, and competences from the local to the national level (center formation).”6 These dual processes gradually shift the focus of politics from the local to the national, and over time voting behavior increasingly homogenizes across all regions, both in any given election as well as in shifts in voting patterns across elections. Thus, for the party system nationalization literature, contemporary regional political distinctiveness is a temporary problem stemming from incomplete modernization/nation-state building. However, other scholars, also drawing from Rokkan and Urwin, posit that geographic remoteness from the center coupled with uneven economic development and the persistence of local cultural identities work to undermine the nationalizing tendencies of modern states. For example, Derek Hearl, Ian Budge, and Bernard Pearson,7 in a study exploring all regions across 11 European states from 1979–1992, found that the centerperiphery cleavages explored by Rokkan and Urwin remained salient across Western Europe and that the most politically distinctive regions appear

Downloaded by [185.120.125.129] at 19:40 04 May 2016

122

B. Cartrite & D. Miodownik

to be associated with cultural, and in particular linguistic, distinctiveness. The authors note as well, however, that regional political distinctiveness is also particularly driven by left-wing voting patterns related to economic development. Explorations of regional political distinctiveness remain divided across a range of areas, as we will shortly see, with few attempts to develop more comprehensive models incorporating the range of factors demonstrated by various literatures as significant. In part, this lack of synthesis can be attributed to the narrow focus on the political fortunes of regional political parties, where such parties exist; this approach fails to capture dynamics of political distinctiveness that may manifest absent such parties. Next, we examine three popular explanations for the persistence of regional political distinctiveness: linguistic distinctiveness, geographic remoteness, and economic conditions. While all three may affect political distinctiveness independently, we note that the bulk of their impact is mediated through their contribution to the appearance and success of regional parties. That said, we argue that linguistic distinctiveness affect patterns of political distinctiveness above and beyond its impact on the emergence and success of regional parties.

Antecedents of Distinctiveness Considerable scholarly research has consistently determined that, where a regional population shares a common language, history, religion, and/or ethnic identity distinct from that of the larger state, such regional distinctiveness is strongly related to the emergence of regional political movements that seek institutional distinctiveness to protect and foster cultural distinctiveness.8 A number of scholars9 have noted the important connection between linguistic distinctiveness and the success of regional parties. Britt Cartrite10 demonstrates that linguistic movements frequently provide both the normative and, often, the organizational underpinnings for regional political-party formation and success. Although the literature differs as to why voters from ethnically or culturally distinct areas ultimately behave differently from voters in other areas, even in the absence of regional political parties, the clear expectation is that cultural distinctiveness translates into distinct political behavior. This is particularly true in democracies, especially in periods immediately following democratization,11 and the most widely used indicator of regional political distinctiveness in this literature is votes for ethnoregionalist parties, most often in national-level elections.12 However, scholars (and policy makers) are divided on the particular role linguistic distinctiveness plays in such movements, arguing, at times, that the impact of cultural difference on distinctiveness is much deeper and

Downloaded by [185.120.125.129] at 19:40 04 May 2016

Determinants of Regional Political Distinctiveness

123

transcends time and space. Anthony Smith13 notes that linguistic distinctiveness has long been considered a fundamental social cleavage demarcating group boundaries. In the 18th century, Johann Gottfried Herder argued that language embeds historical experience and a sense of place to provide a deep sense of community among speakers; as such, languages were to be valued and preserved, a theme advanced by language scholars and activists in the subsequent Romantic period.14 For state builders in the 18th and 19th century, the persistence of local languages presented a fundamental barrier to the program of political unification. For example, in 1794 in a report to the French Republic’s Committee of Public Safety, Bertrand Bar`ere de Viezac argued “federalism and superstition speak low-Breton; emigration and hate for the Revolution speak German; the counterrevolution speaks Italian, and fanaticism speaks Basque,”15 paving the way for initial attempts, via public education, to create a homogenous Parisian French-speaking polity throughout France precisely to undermine local political distinctiveness and unify the nation. As Eugen Weber16 documents in significant detail, by the 1880s public education in France was beginning to realize this outcome, as local dialects and languages began to disappear under the pressure of Parisian French. With the rise of conceptualizations of self-determination in 19th-century Europe and, more clearly, in the wake of World War I, rights for ethnic groups qua language groups were formalized in the “minority treaties” system (although, as Jennifer Jackson Preece17 notes, Western leaders suggested these treaties were intended primarily to preserve peace, during which ongoing homogenization could and should continue). Following World War II, a series of conventions were established to protect languages, and by extension cultures, from extinction. Linguistic preservation came to be understood as a fundamental human right, with a number of international agreements in the 1990s being developed by the United Nations, the Council of Europe, the Organization for Security and Cooperation in Europe, and the European Union to provide for the protection in particular of linguistic minority groups. Anthony Alcock18 argues that signatory states to these instruments, particularly in Western Europe, had begun to resign themselves to the persistence of ethnolinguistic groups, and the potential for regional political activism against the state that might result. Conversely, a number of prominent scholars have expressly rejected the link between language and ethnic distinctiveness on a variety of grounds. Donald Horowitz19 notes that linguistic differences are frequently markers of urban-rural, class, or other divides, while, in some cases, groups speaking the same language nonetheless see themselves as separate, although he notes that in some multilingual states linguistic distinctiveness may become a source for political demands. Indeed, in Smith’s earlier seminal The Ethnic Origins of Nations he argues that “language, long held to be the main, if not the sole, differentiating mark of ethnicity, is often irrelevant or divisive for

Downloaded by [185.120.125.129] at 19:40 04 May 2016

124

B. Cartrite & D. Miodownik

the sense of ethnic community,” noting that, while linguistic distinctiveness may bolster cultural distinctiveness, other cultural elements are often more important; further, the malleability of language itself undermines its relative power.20 Further, some scholars have argued that cultural distinctiveness, whether based in language or not, is in fact insignificant in the generation of regional political movements and political distinctiveness more broadly.21 Such research identifies a range of examples where regional cultural distinctiveness fails to predict the emergence of regional political movements as well as a number of regional political movements whose foundation and impetus do not appear to derive from cultural differentiation, although often such parties may adopt the mantel of advancing the local language or culture. Thus, there is considerable disagreement in the literature as to whether cultural distinctiveness broadly or linguistic distinctiveness in particular drives regional party formation. On the other hand, most scholars cited above would agree that culture in general and language in particular binds social groups, providing them particular (often unique) frames of reference. Such deeply embedded, often implicit particularity, we argue, is manifested in unique patterns of political behavior, independent of the appearance of regional parties, and consequently regional political distinctiveness. To contribute to this debate, we hypothesize the following: The greater the linguistic distinctiveness then the greater the political distinctiveness. Interestingly, although scholars working in the regionalism and centerperiphery literatures note the co-occurrence of regional distinctiveness with physical remoteness from the core, distance itself is rarely taken directly into account in such studies to explain the emergence of regional parties. Scholars such as Stein Rokkan, Karl Deutsch, and Ernest Gellner noted that advances in transportation and communication technologies serve to reduce the political and economic impact of physical remoteness, one of the causal vectors discussed in the nationalization literature above; ongoing innovations in these areas should therefore increase this trend. However, a range of other scholars emphasize that physical remoteness remains a significant factor affecting the emergence and success of regional parties and consequently regional distinctiveness despite such technological innovation. Trevor Corner22 highlights the persistence of cultural distinctiveness, economic underdevelopment, and relative political isolation of remote, particularly insular, regions in Europe despite decades of policies across many European states to reduce such disparities. Indeed, the European Union has specifically targeted remote regions as areas where economic convergence remains the most problematic.23 Thus, the expectation for many scholars in this area is that regional political distinctiveness increases with remoteness from the core. Whereas spatiality is an often-overlooked dimension of regional political distinctiveness, economic distinctiveness has long been one of the primary relationships under consideration in the literature. Some argue that

Downloaded by [185.120.125.129] at 19:40 04 May 2016

Determinants of Regional Political Distinctiveness

125

the greater the economic underdevelopment of peripheral regions the more likely that regional political movements demanding financial policies to address this imbalance emerge.24 However, a number of scholars disagree,25 highlighting the persistence of dependency by peripheral regions on the core, a relationship that undermines the potential for regional political mobilization as potentially threatening much-needed economic support. Peter Gourevitch26 concludes instead that regional political distinctiveness is more likely to be found in areas that are relatively wealthy, both because resources for political mobilization, particularly for regional political parties, are more readily available and as an assertion of demands for political power commensurate with disproportionate economic power. Indeed, Alberto Alesina and Enrico Spolaore27 argue that economic redistribution of resources towards underdeveloped areas serves to stimulate regional political mobilization in wealthier regions, from which such funds are likely to be disproportionately drawn; thus, the combination of economic resources and political grievance may serve as a powerful impetus to the formation and success of regional political parties. Taken together we expect that regional political distinctiveness will be higher in economically distinctive regions. We endeavor in this study to develop an initial model that can address some of these conceptual and empirical shortcomings, graphically illustrated in Figure 1. The figure represents the rationale presented in the discussion above: We anticipate that linguistic differences, geographic peripheriality, and economic status positively impact the formation and success of regional political parties, which in turn contributes to regional political distinctiveness. However, we also expect that regional linguistic distinctiveness will have an independent direct effect on political distinctiveness.

FIGURE 1 Theoretical model.

126

B. Cartrite & D. Miodownik

Downloaded by [185.120.125.129] at 19:40 04 May 2016

RESEARCH DESIGN This study provides an exploratory evaluation of the relative importance of linguistic distinctiveness, geographic remoteness, relative economic performance, and regional political parties in generating regional political distinctiveness. In particular, this research takes an approach similar to that of Hearl, Budge, and Pearson by focusing on the same 11 countries for the analysis.28 However, in comparison to that earlier work, in this study the concept of the “region” is more clearly articulated and systematically applied, the index developed to measure political distinctiveness more closely resembles similar statistical measures and incorporates voter turnout in the index (rather than treating it as a variable independent of regional political distinctiveness), the dummy code for regional linguistic distinctiveness is replaced by a continuous measure, and regional political parties are addressed as an independent predictor of regional distinctiveness. The research design discussion below begins with the dependent variable, regional political distinctiveness, including the processes for its calculation, followed by the mediating variable of regional political parties. Finally, the independent variables of linguistic distinctiveness, economic distinctiveness, and distance are examined in turn.

Dependent Variable: Political Distinctiveness Index This study develops a Political Distinctiveness Index (PDI) based on all national lower house legislative elections since 1990 (see Appendix Table 1 for the complete list of elections included in the dataset) to quantify the TABLE 1 List of “Regions”

Belgium (BE) Denmark (DK) France (FR) Germany (DE) Greece (EL/GR) Ireland (IE) Italy (IT) Netherlands (NL) Portugal (PT) Spain (ES) United Kingdom (UK)

Constituency Level

Number

NUTS level

Region County Region State Region Regional Authority Region (Province) Province District / Autonomous Region Autonomous Community/City English Region / Scotland / Wales / Northern Ireland

3 17 22 16 13 8 21 12 20 19 12

1 3 (pre-2006) 2 1 2 3 2(3)1 2 n/a 2 1

1 The Italian region of Trentino-Alto Adige/S¨ udtriol is unique among Italian regions in that it does not have a directly elected assembly for the entire region; rather, the regional assembly is comprised of the representatives of the provincial assemblies of Trento and South Tyrol. Consequently, this study codes the two provinces as separate regions.

Downloaded by [185.120.125.129] at 19:40 04 May 2016

Determinants of Regional Political Distinctiveness

127

relative distinctiveness of voting behavior at the regional level. Election results were obtained from a variety of online election databases, online statistical resources from various governments, and French elections data provided directly by agreement with SciencesPo.29 National, lower house election data were collected, using first-round (France) or multimember district (Germany: “Zweistimmen”) results where multiple voting options existed. In all cases, raw tallies rather than datasets that aggregated small parties into larger (typically “other”) categories were used in order to preserve the granularity of the original electoral data and to capture differences in voting behavior for very small parties. As noted above, the party systems nationalization literature examines variations in political behavior across a variety of dimensions; this has produced an intense debate as to the most appropriate measures and their relative strengths and weaknesses.30 While most of the measures developed through this debate are not appropriate for this study, given their focus on individual party nationalization, party systems at the national level, linkages across levels of governance, or dynamic changes over time, our focus on identifying territorial distinctiveness within countries is most similar to studies that use an index of dissimiliarity (DIS),31 where differences in voter support for each party in each region compared to average support across all regions are calculated, summed, and then divided by two to avoid doublecounting:  1   VP − V¯ P  . 2 P =1 n

DIS =

(1)

However, as Caramani and Bochsler each note, this approach effectively under weighs larger deviations from the mean, thereby dampening the calculation of distinctiveness.32 In addition, all studies in this area measure distinctiveness as the relative share of valid votes cast, rather than as the share of the electorate, thereby omitting voter turnout from the conceptualization of voting distinctiveness. While some studies, such as Hearl et al.,33 include voter turnout as an independent variable causing political distinctiveness, even they ultimately reject the inclusion of turnout in their model as both too weak and too problematic for inclusion, in part because voter abstention can reasonably stem from voter apathy, protest, or satisfaction with the status quo, each of which would impact the understanding of overall voting distinctiveness in different ways. To address these two issues, we develop the political distinctiveness index (PDI). Whereas DIS simply sums the absolute value of the difference in vote share within a given region from the average across all regions for each party, PDI reflects the statistical measure of standard deviation by squaring the differences in order to emphasize larger deviations from the mean, sum those squared differences and then take the square root of the sum, in line

128

B. Cartrite & D. Miodownik

Downloaded by [185.120.125.129] at 19:40 04 May 2016

with what Morgenstern et al.34 term “static standard deviation” measures. In addition, rather than dividing the number of votes by valid votes, we divide by the size of the electorate, which therefore captures variations in voter turnout within the index, although, as a consequence, it reduces the differences in the share of votes each party receives, thus making the measure relatively more conservative in assessing distinctiveness.35 Finally, we then take the average PDI score for each region across all national elections in the dataset.36 Given the unique use of electorate, rather than valid votes, in the calculation of PDI, we also include analyses using the more common DIS index as well as PDI2, which retains the squaring of differences but calculates variance on the basis of valid votes.   n   2 VP − V¯ P . PDI = 

(2)

P =1

As Morgenstern et al.37 note, there is no single perfect measure for party nationalization, as each requires trade-offs relating to the research design; for any given research, however, some measures may be preferable to others. PDI has a number of advantages as a measure of relative political distinctiveness. First, the measure has a theoretical range from near 0 to approaching 1. Zero would indicate that voter turnout and votes for parties in a region perfectly match averages across all regions, a highly unlikely occurrence. A PDI of 1 would indicate perfect divergence from the average support across all parties; because the votes within a region are included in the calculation of party-support averages, the actual upper limit is a function of the number of regions in an election, with the higher the number of regions the closer to 1 a region can theoretically achieve. A second distinct advantage to the measure is that, while the existence of parties that only contest elections in part of a country, or indeed a single region, necessarily increases the distinctiveness of those areas relative to others, by the same process, regions without the regional party also have an increase in PDI given that the average support across all regions for the regional party is necessarily greater than 0. In this way, the impact of a regional party on distinctiveness is not simply for the region(s) where it contests elections but also influences the distinctiveness of all regions. A third advantage lies in using party names rather than attempting to aggregate parties along other dimensions, such as ideology. Simply relying on party names avoids the potential for significant errors in attempting to assign to specific parties particular agendas or political positions and then assessing distinctiveness on such bases. In cases where a party contests elections in different constituencies with slightly different spellings (France in particular), the different spellings are corrected to the most common;

Downloaded by [185.120.125.129] at 19:40 04 May 2016

Determinants of Regional Political Distinctiveness

129

where small parties jointly contest elections with larger parties as blocks or a coalition, we code votes as belonging to the national or largest party. We measure political distinctiveness at the regional level in 11 European countries: Belgium, Denmark, France, Germany, Greece, Ireland, Italy, the Netherlands, Portugal, Spain, and the United Kingdom.38 For this study, the “region” is defined as the highest subnational political unit in each case, which allows us to be consistent across the 11 countries in terms of the hierarchical location of the “regions” within the state structure rather than determining a different institutional level for each of the 11 states. In most cases, these units correspond to a Eurostat Nomenclature of Territorial Units for Statistics (NUTS) categorization; this approach provides for considerable flexibility in data aggregation across different administrative levels, as subordinate NUTS regions are entirely contained within their superordinate region.39 Unfortunately, the use and meaning of NUTS regions varies considerably across European states, and many do not use all NUTS levels or in the same manner. Fortunately, for most countries in the dataset, electoral constituencies correspond to NUTS administrative regions and data were collected at that level. In Belgium and the Netherlands, multimember electoral constituencies are equivalent to NUTS 2, while in Germany the proportional multimember constituencies are aggregated by federal state (Lander, NUTS 1). In the ¨ case of Denmark, the old counties (amt) corresponding to older NUTS 3 districts were abolished in 2007 and new constituencies and NUTS 3 boundaries drawn; in this case, the newer constituencies have been located within the older boundaries to allow for the use of historical data. In Ireland and the United Kingdom, electoral constituencies are much smaller than the administrative units, requiring that individual constituencies be assigned to NUTS regions; where electoral boundaries cross administrative boundaries, assignment of the constituency was determined based on the location of the majority of the population of the district. For geographic information system mapping purposes, the latest NUTS 3 maps (2010) made available by Eurostat were used where possible, supplemented with a separate layer file for Portugal and Denmark reflecting the different regional boundaries.40 Table 1 summarizes the regions by country used in this study. Although the 161 regions in this study share the feature of being the highest political subdivision in each state, there is considerable diversity across the cases. In terms of population (M = 2,398,309; Mdn = 1,262,489; SD = 2,908,613), the largest region is Nordrhein-Westfalen (17,848,113) in Germany, while the smallest region is the island of Bornholm (43,347) in Denmark. The area of the regions is likewise very diverse (M = 14,745 km2; Mdn = 9,398 km2; SD = 16,394 km2), with Spain hosting both the largest ´ (94,226.1 km2) and the smallest regions of the region in Castilla y Leon cities of Melilla (13.4 km2) and Ceuta (19.4 km2). Perhaps not surprisingly, therefore, population density across all regions is likewise quite diverse

Downloaded by [185.120.125.129] at 19:40 04 May 2016

130

B. Cartrite & D. Miodownik

FIGURE 2 Standardized average regional political distinctiveness index scores.

(M = 424.80; Mdn = 128.65; SD = 1078.08), ranging from the Belgian capital of Brussels (7295.8) to Beja in Portugal (14.94). Figure 2 provides a visualization of political distinctiveness across all 161 regions, shaded based on the number of standard deviations above or below the mean across all regions. Not surprisingly, although no regions have a PDI = 0, the distribution of values across the 161 regions is such that most regions are only of modest political distinctiveness. Average PDI scores range from 0.0279 (Gelderland, Netherlands) to 0.449 (Valle d’Aosta, Italy) (M = 0.093; Mdn = 0.077; SD = 0.065). Of the 161 cases, 91 (56.5%) are within one-half standard deviation of the mean and 147 (91.3%) are within one standard deviation of the mean. Conversely, three regions are more than five standard deviations above the mean: Valle d’Aosta; South Tyrol (PDI = 0.4349); and Northern Ireland (PDI = 0.4299). Table 2 lists the 13 regions with average PDI scores at least one standard deviation above the mean, sorted from highest to lowest average PDI. The list captures many of the “usual suspects” of regions studied for their ethnic and political distinctiveness, including Northern Ireland, the Flemish Region, Catalu˜na, and Scotland. With the partial exception of Northern Ireland and depending on how one thinks of linguistic distinctiveness in Belgium, most of the regions on the list, including most of the top eight, are notable for having regional languages distinct from the national language. However,

Determinants of Regional Political Distinctiveness

131

Downloaded by [185.120.125.129] at 19:40 04 May 2016

TABLE 2 Regions with PDI Scores One Standard Deviation Above The Mean or Greater. Country

Region Code

Italy Italy United Kingdom Belgium Spain Spain Belgium France Spain Portugal United Kingdom United Kingdom France

ITC2 ITD1 UKN BE2 ES21 ES51 BE3 FR83 ES70 PTx2 UKM UKK FR63

Region Name Valle d’Aosta South Tyrol Northern Ireland Flemish Region Pa´ıs Vasco Catalu˜na Walloon Region Corse Canarias Beja Scotland South West, England Limousin

Average PDI

Z Score

0.4492 0.4349 0.4299 0.2943 0.2916 0.2722 0.2571 0.1752 0.1744 0.1649 0.1608 0.1591 0.1586

5.478 5.258 5.182 3.094 3.054 2.755 2.522 1.262 1.250 1.104 1.041 1.014 1.007

there are other regions notable for linguistic and political distinctiveness that fail to fall more than one standard deviation above the mean, such as Wales (Z = 0.43), Brittany (Z = −0.13), and Galicia (Z = 0.10), while lesser-known cases make the list, suggesting that regions with strong regionalist political behavior may in fact be overlooked by the broader qualitative literature.

Mediating Variable: Regional Political Parties As noted above, the formation and electoral support for regional political parties represents one of the primary areas of scholarship in the study of regionalism in Europe. Given that most European states use some form of proportional representation for at least a portion of the national lower house seats, small regional parties often have the real possibility of sending candidates to the national legislature. Although “regional” parties vary widely in terms of their underlying ideological perspectives and specific goals, such parties by definition seek to redefine the institutional relationship between the region and the central state41; as such, voter support for regional parties in competitive elections can be understood as a proxy for regionalist sentiment. Here we use a categorical regional party index, coding regions as “0” for no regional party, “1” for a regional party contesting only subnational elections, “2” for a regional party contesting at least two national elections but failing to realize more than 3% of the regional vote, “3” for a regional party exceeding 3% of the regional vote but failing to achieve seats in the national legislature, and “4” for regional parties exceeding 3% and winning seats in at least two elections42 this approach allows for assessing parties in terms of the level at which they contest elections, their electoral support, and their electoral success in a single measure, while a measure based only on the share of votes won by regional parties would not.43 Appendix Table 2 lists the 25 (15.53%)

132

B. Cartrite & D. Miodownik

Downloaded by [185.120.125.129] at 19:40 04 May 2016

regions and regional party names where RegParty > 0; such regions are only found in the following: Belgium (all three regions); Italy (12/22 regions); Spain (7/18 regions); and the United Kingdom (3/12 regions).44 Regional parties contesting national elections clearly affect the PDI score. In fact, their presence increases not only the political distinctiveness of the region(s) in which such parties contest elections but increase the distinctiveness of all regions in a country: Because the average share of the electorate for such parties across all regions, thus partially reducing the distinctiveness of the region(s) in question, all other regions also vary from the mean share of the electorate and, therefore, have their distinctiveness (usually marginally) increased as well. However, the impact of regional parties on PDI is not strictly determined: Given that PDI incorporates not only variation in votes for national parties but also turnout, regions without regional parties can nonetheless be quite politically distinct.

Independent Variable: Linguistic Distinctiveness Categorically assigning a historic language to a region, particularly as a proxy for ethnic distinctiveness, is problematic in some ways, in part precisely due to the homogenizing processes of state building that serves not only to undermine local language usage but to facilitate in- and out-migration into such peripheral regions, such that the population of all regions is mixed, and local language usage (and the politicization of language) results from both regional- and state-level dynamics. This mixing of populations and language usage may serve to weaken local distinctiveness or may in fact generate both cultural and political resistance to perceived cultural or linguistic loss; thus, the number of people with regional language proficiency and estimates of the size of this group could be the result of processes of mobilization and accompanying politicization of language differences against the state rather than simply a direct result of state-building.45 Consequently, rather than attempting to assess local language usage as an explanatory variable for political distinctiveness, this study instead adopts the approach developed by linguists who adopt an approach similar to the phylogenic “tree” of species used in biology, in which assessments are made as to the relationships between existing languages and the linguistic (and temporal) distance back to a common linguistic node. As in biology, such an approach allows linguists to evaluate the relative similarity and distinctiveness of languages by identifying where on a particular “branch” two languages share a common node. Also as in biology, the number of identified “nodes” for a given language varies, as some historical languages have undergone more divisions and subdivisions resulting in new languages than have others. The 2014 Ethnologue catalog provides the linguistic address for over 7000 known living languages, including all state and regional languages in

Downloaded by [185.120.125.129] at 19:40 04 May 2016

Determinants of Regional Political Distinctiveness

133

Western Europe46; the information on each language includes descriptions of regional locations where the languages are in use, allowing for the assignment of languages to all regions under consideration here. To generate the Linguistic Distinctiveness (LD) measure, this study begins by identifying the “state” languages associated with each of the 11 countries,47 specifying the full language “address” and counting the number of nodes.48 For example, French is classified as “Indo-European, Italic, Romance, Italo-Western, Western, Gallo-Iberian, Gallo-Romance, Gallo-Rhaetian, O¨ıl, French” and having 10 nodes. Regions where there is an associated local autochthonous language are also identified and the language address of the most distinctive local language is determined (for example, in Midi-Pyrenees the local language of Occitan is identified as “Indo-European, Italic, Romance, ItaloWestern, Western, Gallo-Iberian, Ibero-Romance, Oc, Occitan.” The index is calculated as the following: LD = 1 − (Nc /Ns ),

(3)

where Nc is the number of language address nodes the regional language has in common with the state language and Ns is the total number of nodes in the state-language address.49 By standardizing distinctiveness by the length of the state language, the index represents a continuous measure from 0 to 1 for all regions in the study, with 0 representing no linguistic distinctiveness between the region and the state and 1 representing cases in which the regional and the state language have no nodes in common; in the dataset, only regions where Basque is a local language achieve the maximum score of LD = 1, as Basque is the only non-Indo-European regional language remaining in our dataset and, therefore, has no nodes in common with any state language.50 Thus, using the example of the Midi-Pyrenees region for the Occitan language: LD = 1 − (6 [nodes in common]/10 [nodes for French]) = 0.40.

(4)

Of the 161 regions in the dataset, 60 (37.3%) are coded as having a regional language distinct from the state language (M = 0.171; Mdn = 0; SD = 0.287). Appendix Table 3 lists all of the linguistically distinct regions with the name of the local language and the LD score.

Independent Variable: Relative Economic Distinctiveness To assess the impact of economic distinctiveness, we calcualte relative gross domestic product (GDP) per capita, dividing regional GDP (RGDP) per capita in 2013 by national GDP per capita in 2013. The measure allows us to take into account both regions that are poor relative to the national average as well as those that are better off. RGDP ranges from a low of 0.5789 (Border

134

B. Cartrite & D. Miodownik

TABLE 3 Standardized Coefficients by Dependent Variable Measure.

Downloaded by [185.120.125.129] at 19:40 04 May 2016

Regional Party Distance Economy Language Error 1 Distinctiveness Regional Party Language Error 2 Chi2 R2 TLI RMSEA N +

PDI

PDI2

DIS

PDI+

PDI2+

DIS+

PDI++

.375∗∗∗ .242∗∗∗ .382∗∗∗ .779∗∗∗

.375∗∗∗ .242∗∗∗ .382∗∗∗ .779∗∗∗

.375∗∗∗ .242∗∗∗ .382∗∗∗ .779∗∗∗

.387∗∗∗ .235∗∗∗ .351∗∗∗ .797∗∗∗

.387∗∗∗ .235∗∗∗ .351∗∗∗ .797∗∗∗

.387∗∗∗ .235∗∗∗ .351∗∗∗ .797∗∗∗

.426∗∗∗ .200∗∗ .332∗∗∗ .788

.487∗∗∗ .126∗ .831∗∗∗ .668∧ .310 1.049 .000 161

.438∗∗∗ .155∗∗ .849∗∗∗ 1.671∧ .279 1.013 .000 161

.464∗∗∗ .188∗∗ .818∗∗∗ .490∧ .331 1.053 .000 161

.485∗∗∗ .110∧ .841∗∗∗ .523∧ .293 1.060 .000 159

.436∗∗∗ .137∗∗ .861∗ 1.768∧ .259 1.010 .000 159

.462∗∗∗ .185∗∗ .825∗∗∗ .490∧ .331 1.053 .000 159

.381∗∗∗ .191∗∗ .870∗∗∗ 2.152∧ .243 .994 .022 161

Basque country and Navarra removed from analysis. ∗ ∗∗ p > .1. p < .1. p < .05. ∧p < .001.

∗∗∗

++ Regional party treated as dummy variable.

Regions, Ireland) to 1.841 (Brussels, Belgium) (M = 0.938, Mdn = 0.876, SD = 0.232). In the case of Portugal, where NUTS regions do not correspond to electoral districts, we assigned regional GDP based on the superordinate NUTS 2 regions; thus, some regions are coded as having the same relative GDP. In the case of Denmark, we similarly applied superordinate data to the regions, in this case using economic data based on the newer, larger, NUTS regions.

Independent Variable: Distance Finally, to evaluate the relative impact of distance on political distinctiveness, we include in the analysis a simple measure of linear distance in kilometers from the national to the regional capital for each region in the dataset based on estimates provided by Google Maps. Values range from the state capital regions (0) to the Canary Islands of Spain (1769 km) (M = 269.59; Mdn = 225; SD = 235.5).

ANALYSIS Bivariate Models We begin be exploring the independent association between each of the exogenous variables: distance, economic strength, and language differences tranformed into dummy variables and a dummy variable indicating the presence of a successful regional party. Figure 3 plots results of a mean comparison between the independent indicators and regional party, attesting to a

Downloaded by [185.120.125.129] at 19:40 04 May 2016

Determinants of Regional Political Distinctiveness

135

FIGURE 3 Regional party probability: (A) distance to capital; (B) economic strength; (C) linguistic differences.

positive and statistically significant association between each of the indicators and the regional party success. Panel A compares the appearace of parties in regions located close to the capital (distance < 160 km) coded 0 to distant regions (distance ≥160 km) coded 1. We chose 160 km, as it represents the cutting pont for the first quartile (note that the choice of other “cutting points” did not change the analysis substantively). As such, the figure evidences that 75% of the regions in our data, regions coded as distant, are approximatly three times more likely to have succesful regional parties in comparison to regions located closer to the capital, F (Distance) = 10.513, p < .01. Panel B compares affluent regions (regions with a local GDP per capita larger than the national mean) coded 1 to economically poor regions (regions with GDP per capita smaller than the national mean) coded 0. The figure indicates that regional parties are about 40% more likely in regions where the local economy is stronger than the national mean, F (Economy) = 6.571, p < .05. Lastly, panel C compares the likelihood of regional party success in

Downloaded by [185.120.125.129] at 19:40 04 May 2016

136

B. Cartrite & D. Miodownik

regions with a linguistic heritage different than the center’s (LD > 0; coded 1) and regions with no such differences (coded 0). Note that the dummy variable does not capture either the extent of the language differences or the number of people that can speak or understand the regional language. The mean comparison indicates that linguistically distinct regions are about four times more likely to experience successful regional parties than regions without an unique linguistic heritage than the center, F (Language) = 50.403, p < .0001. Next, we examine the effect of the three exogenous variables and the mediator (regional party) on levels of political regional distinctiveness. Panels A, B, and C suggest that distance from the capital seems not to increase the overall level of voting distinctiveness, that economic might is weakly but significantly (in conventional statistical terms) linked to political distinctiveness, and that linguistic uniqueness contributes to large differences in political distinctiveness scores, F (Distance) = 0.922; F (Economy) = 5.071; and F (Language) = 13.036, p < .0001, respectively. Panel D evidences that regional party success constitutes a strong predictor of levels of regional political distinctiveness, as the average PDI score is almost twice as large for regions with successful regional parties when compared to regions with no (or only marginal) regional parties, F (Regional Party) = 42.920, p < .0001. Taken together Figures 3 and 4 provide initial indications supporting the expectation that the effects of geographic, economic, and linguistic status on regional political distinctiveness is mediated by the appearance and success of regional parties, but that linguistic differences, that is, our proxy for a unique cultural/historic background, may affect voting distinctiveness independently of its effect on the emergence of regional parties. The multivariate model discussed next provides additional support for these expectations.

Multivariate Models To test our expectation, we ran a “path analysis” using the AMOS package for structural equation modeling (SEM).51 In essence, path analysis, a subset of SEM, extends multivariate linear regression by allowing the simultaneous estimation of a series of multiple regressions. In our case, we use this technique to estimate simultaneously the effect of language differences, geographic distance, and economic status on the emergence and success of a regional party, and the combined effect of regional party and language differences on the extent of political regional distinctiveness. This approach is preferred to a more conventional linear regression in which we might include interaction terms for, in this case, the independent variables and regional parties, as such an operationalization would knowingly wrongly specify the relationships of the different factors and not allow us to assess the causal impact of language, economics, and distance on regional parties.

Downloaded by [185.120.125.129] at 19:40 04 May 2016

Determinants of Regional Political Distinctiveness

137

FIGURE 4 Regional political distinctiveness (mean): (A) distance to capital; (B) economic strength; (C) linguistic differences; (D) regional party.

Path analysis models produce sets of statistics compatible to those generated by a typical linear regression model. Figure 5 shows results of the simultaneous multivariate analysis,52 including linguistic distinctiveness as having a direct effect on PDI.53 The various “goodness of fit” statistics indicate that the model fits the data quite well.54 With confidence in the fit of the model, we can move to describe the actual results depicted in Figure 5. The analysis suggests that all three independent predictors have a statistically significant positive effect on the emergence and success of regional parties. From the three predictors tested, language differences have the strongest effect on the appearance and success of regional parties (β = 0.382, p < .001), followed by the distance of the region from the capital (β = 0.375, p < .001), and economic difference (β = 0.242, p < .001) (in other words, economically distinct and distant regions are more likely to experience the emergence and success of regional parties). Note, however, that the three indicators explain only 39% of the variation in the emergence and success of regional parties. This is also evident from the

Downloaded by [185.120.125.129] at 19:40 04 May 2016

138

B. Cartrite & D. Miodownik

FIGURE 5 Multivariate analysis – language.

large standardized coefficient of the error term (β = 0.779, p < .001) more than twice as large as the effect of language. Moving to political distinctiveness, Figure 5 indicates, obviously, that the emergence and success of regional parties affects the political distinctiveness score (β = 0.487, p < .001), but it also provides sufficient indication in support of the expectation that the linguistic differences, our proxy for the cultural uniqueness of a region, has a substantive independent effect on regional political distinctiveness (β = 0.126, p < .1). The marginal statistical significance (in relation to the .05 convention) should probably be attributed to the relative small number of observations (N = 161). The R 2 value suggests

Downloaded by [185.120.125.129] at 19:40 04 May 2016

Determinants of Regional Political Distinctiveness

139

that the combination of regional party and language distinctiveness explains about 31% of the variation in regional political distinctiveness, while the remaining variation is explained by unobserved predictors captured by the large coefficient of error term (β = 0.831, p < .001). Table 3 summarizes the various versions of the model.55 PDI2 differs from PDI by using each party’s share of valid votes cast, rather than the electorate, in the calculation, while DIS is the more conventional measure of dissimilarity discussed above. We reran all three models excluding the Basque regions (indicated with a +) to check for the robustness of the results absent those linguistically outlier regions; in addition, we reran the analysis using a simple dummy variable for the presence of a regional party (indicated by ++). Interestingly, PDI appears to be a relatively conservative estimator of distinctiveness, yielding lower coefficient scores than the other measures, indicating that variances in turnout have a dampening effect on the relative distinctiveness of voting patterns. The direct effect of language is evident, albeit weakly, against PDI, whereas its impact is much more significant for PDI2 and DIS. While the exclusion of the Basque regions weakens the impact of language to a notable degree (particularly for PDI+), the direct effect of language remains statistically significant by the alternative measures.

CONCLUSION This study represents a considerable improvement on earlier studies for a number of reasons. First, the PDI index developed here permits the inclusion of all regions from each case within the analysis, improving upon earlier similar measures and allowing for a more comprehensive assessment of the impacts of different factors across all 161 regions. Second, the path analysis presented here allows for the treatment of regional political parties as factor driven, in part, by linguistic distinctiveness, economic distinctiveness, and distance but simultaneously contributing to overall regional political distinctiveness, rather than equating political distinctiveness with the appearance of regional parties. Third, we demonstrate that, while the effects of distance are fully mediated and economic distinctiveness almost fully mediated through regional political parties, linguistic distinctiveness has an independent impact on political distinctiveness even absent such regional parties, and that these results are even more prominent using measures based on valid votes. Fourth, we show that the degree of linguistic distinctiveness, and not the mere presence of a regional language, drives political distinctiveness. The continuing emphasis on linguistic rights and the revitalization of regional languages can therefore be expected to increase regional political distinctiveness in those areas, laying the foundations for the formation of regional political parties that then draw upon remoteness and economic distinctiveness to contest political elections.

Downloaded by [185.120.125.129] at 19:40 04 May 2016

140

B. Cartrite & D. Miodownik

There are a number of future lines of research following from this study. First, while this study uses average PDI across all elections since 1990, a more developed dataset with economic measures from earlier years would allow for an analysis of the evolution of PDI over time as well as across regions to more clearly establish the trajectory of regionalism over time in order to better assess the expectations of the party systems nationalization literature. A second line would be to expand the number of countries included in the study to capture more regions and, therefore, to expand the generalizability of the research. Given the prominence of economic explanations for regional distinctiveness in the literature, a more robust analysis of the impacts of different aspects of economic distinctiveness on political distinctiveness seems promising. Finally, a more nuanced exploration of the growth and impact of regional political parties in different political contexts would better clarify the mediating effect of such parties on regional political distinctiveness. Thus, this study is perhaps merely the next step towards a more comprehensive, robust, and integrative evaluation of the many dynamics explored by scholars of regionalism for the past half century.

ACKNOWLEDGMENTS The authors would like to thank participants of the International Political Science Association RC 14 (Ethnicity and Politics) meetings in Sydney (2013), Montreal (2014), and Edinburgh (2015) and those at the Midwestern Political Science Association meetings (2014, 2015) for feedback on this project at various stages. We also thank the anonymous reviewers for their helpful and constructive feedback. Finally, Dr. Cartrite would like to thank the students of his POL 215 course (2014), and, in particular, Luke Ashton, Jeremy Johnson, and Alexander Sprague for their contributions and frequent critiques of this project.

NOTES 1. Arjan Schakel, “Nationalisation of Multilevel Party Systems: A Conceptual and Empirical Analysis,” European Journal of Political Research 52(2): 212–236 (2013). 2. Elmer Schattschneider, The Semisovereign People (New York: Holt Rinehart and Winston, 1960). 3. Daniele Caramani, The Nationalization of Politics: The Formation of National Electorates and Party Systems in Western Europe (Cambridge MA: Cambridge University Press, 2004). 4. Albert O. Hirschman, Exit, Voice, and Loyalty: Responses to Decline in Firms, Organizations, and States (Cambridge MA: Harvard University Press, 1970). 5. Stein Rokkan and Derek Urwin, eds., The Politics of Territorial Identity: Studies in European Regionalism (London: Sage, 1982). 6. Caramani, The Nationalization of Politics, 32. 7. Derek Hearl, Ian Budge, and Bernard Pearson, “Distinctiveness of Regional Voting: A Comparative Analysis across the European Community (1979–1993),” Electoral Studies 15(2): 167–182 (1996). 8. For example, Lieven De Winter and Huri T¨ursan, eds., Regionalist Parties in Western Europe (New York: Routledge, 1998); Peter Gourevitch, “The Reemergence of ‘Peripheral Nationalisms’: Some

Downloaded by [185.120.125.129] at 19:40 04 May 2016

Determinants of Regional Political Distinctiveness

141

Comparative Speculations on the Spatial Distribution of Political Leadership and Economic Growth,” Comparative Studies in Society and History 21(3): 302–322, (1979). Michael Hechter, Internal Colonialism: The Celtic Fringe in British National Development (Berkeley: University of California Press, 1975). Stein Rokkan and Derek Urwin, Economy, Territory, Identity: Politics of West European Peripheries (London: Sage Publications, 1983). 9. For example, John Ishiyama and Marijke Breuning, Ethnopolitics in the New Europe (Boulder, CO: Lynne Rienner, 1998); Miroslav Hroch, Social Preconditions of National Revival in Europe: A Comparative Analysis of Patriotic Groups among the Smaller European Nations (New York: Columbia University Press, 1985). 10. Britt Cartrite, “Reclaiming Their Shadow: Ethnopolitical Mobilization in Consolidated Democracies” (doctoral dissertation, University of Colorado, Boulder, 2003). 11. J´ohanna Birnir, Ethnicity and Electoral Politics (Cambridge: Cambridge University Press, 2006). 12. For example, Derek Urwin, “Harbinger, Fossil, or Fleabite?: ‘Regionalism’ in the West European Party Mosaic,” in Hans Daalder and Peter Mair, eds., Western European Party Systems: Continuity & Change (London: Sage, 1983), 221–256; Lieven De Winter and Huri Tursan (eds.), Regionalist Parties in Western Europe (Routledge: New York, 1998). Eve Hepburn, “Introduction: Reconceptualizing Sub-state Mobilization,” Regional & Federal Studies 19(4–5): 477–499 (2009). 13. Anthony Smith, “The Power of Ethnic Traditions in the Modern World,” in Steven Grosby and Athena Leoussi, eds., Nationalism and Ethnosymbolism: History, Culture and Ethnicity in the Formation of Nations (Edinburgh: Edinburgh University Press, 2007), 325–336. 14. Patrick Geary, The Myth of Nations: The Medieval Origins of Europe (Princeton: Princeton University Press, 2003). 15. Quoted in James Jacob and David Gordon, “National Minority Policy in France,” in William R. Beer and James E. Jacob, eds., National Minority Policy and National Unity (Totowa NJ: Rowman & Allanheld, 1985), 114. 16. Eugen Weber, Peasants into Frenchmen: The Modernization of Rural France 1870–1914 (Stanford CA: Stanford University Press, 1976). 17. Jennifer Jackson Preece, National Minorities and the European Nation-States System (Oxford, UK: Clarendon Press, 1998), 88–89. 18. Anthony Alcock, A History of the Protection of Regional Cultural Minorities in Europe: From the Edict of Nantes to the Present Day (New York, NY: St. Martin’s Press, LLC, 2000). 19. Donald Horowitz, Ethnic Groups in Conflict (Berkeley: University of California Press, 2000). 20. Anthony Smith, The Ethnic Origins of Nations (Cambridge MA: Blackwell, 1986), 27. 21. For example, James Fearon and Pieter van Houten, “The Politicization of Cultural and Economic Differences: A Return to the Theory of Regional Autonomy Movements,” paper presented at the Fifth Meeting of the Laboratory in Comparative Ethnic Processes (LiCEP), 10-11 May, Stanford University (2002); Pieter van Houten “Regional Assertiveness in Western Europe: Political Constraints and the Role of Party Competition” (doctoral dissertation, Chicago: University of Chicago, 2000). 22. Trevor Corner, “The Maritime and Border Regions of Western Europe,” Comparative Education 24(2): 303–322 (1988). 23. See ESPON 2013 Programme: The Development of the Islands – European Islands and Cohesion Policy, http://www.espon.eu/export/sites/default/Documents/Projects/ TargetedAnalyses/EUROISLANDS/FinalReport_foreword_CU-16-11-2011.pdf (accessed 15 March 2015). 24. For example, Michael Hechter, Internal Colonialism: the Celtic Fringe in British National Development, 1536–1966 (Berkeley: University of California Press, 1975). 25. For example, Godfrey Baldacchino and Eve Hepburn, “A Different Appetite for Sovereignty?: Independence Movements in Subnational Island Jurisdictions,” Commonwealth and Comparative Politics 50(4): 589–602 (2012); Gert Oostindie, “Dependence and Autonomy in Sub-National Island Jurisdictions: The Case of the Kingdom of the Netherlands,” The Round Table: The Commonwealth Journal of International Affairs 95(386): 609–626 (2006). 26. Peter Gourevitch, “The Re-Emergence of “Peripheral Nationalisms”: Some Comparative Speculations on the Spatial Distribution of Political Leadership and Economic Growth,” Comparative Studies in Society and History 21(3): 303–322 (1979). 27. Alberto Alesina and Enrico Spolaore, The Size of Nations (Cambridge: MIT Press, 2003). 28. Hearl et al., “Distinctiveness of Regional Voting,” 171–174.

Downloaded by [185.120.125.129] at 19:40 04 May 2016

142

B. Cartrite & D. Miodownik

´ 29. Manuel Alvarez-Rivera, Election Resources on the Internet, http://www.electionresources.org/ (accessed 15 March 2015); Dawn Brancati, Global Elections Database [computer file] (New York: Global Elections Database [distributor]), http://www.globalelectionsdatabase.com (accessed 26 March, 2014); European Elections Database, http://www.nsd.uib.no/european_election_database/ (accessed 15 March 2015); Ken Kollman, Allen Hicken, Daniele Caramani, and David Backer, ConstituencyLevel Elections Archive (Ann Arbor, MI: Center for Political Studies, University of Michigan), http://www.electiondataarchive.org/ (accessed 15 March 2015); Elections legislatives, 1e tour, parcirconscription [fichier e´ lectronique], Minist`ere de l’Int´erieur, CDSP [producers], Centre de Donn´ees Sociopolitiques (CDSP) [distributor]. 30. For extended comparisons, see Daniel Bochsler, “Measuring Party Nationalisation: A New Gini-based Indicator that Corrects for the Number of Units,” Electoral Studies 29: 155–168 (2010); and Scott Morgenstern, John Polga-Hecimovich, and Peter Siavelis, “Seven Imperatives for Improving the Measurement of Party Nationalization with Evidence from Chile,” Electoral Studies 33: 186–199 (2014). 31. Schakel, “Nationalisation of Multilevel Party Systems,” 7; Hearl et al., “Distinctiveness of Regional Voting,” 169; Adrian Lee, “The Persistence of Difference: Electoral Change in Cornwall” (paper presented at the Political Studies Association Conference, Plymouth, UK, 1988). 32. Daniele Caramani, The Nationalization of Politics, 60–63; Daniel Bochsler, “Measuring Party Nationalisation,” 158. 33. Hearl et al., “Distinctiveness of Regional Voting,” 178. 34. Morgenstern et al., “Seven Imperatives,” 188. 35. Note that we do not weight votes by party size, which has the effect of significantly increasing the impact of small parties, nor do we weight regions by population. For a more complete assessment of the impacts of various weighting schema, see Britt Cartrite, Dan Miodownik, Karalyn Nic, and Victoria Bishop, “Comparing Regional Measures of Voting Distinctiveness” (paper presented at the Midwest Political Science Association meeting, Chicago, IL, 8–10 April 2013). 36. Although average PDI are calculated for all regions across time, all other measures are treated as constants for each region. While this necessarily precludes an evaluation of trajectories over time, we are able to assess variation across cases in the other variables on the patterns of distinctiveness over space. 37. Morgenstern et al., “Seven Imperatives,” 190. 38. These states plus Luxembourg were the 12 member states of the European Union in 1990. Luxembourg has no effective regional institutions and is therefore excluded from this study. 39. NUTS 0 are European states. NUTS 1 are the highest subnational administrative level (states in some countries, administrative regional groups in others). NUTS 2 are usually administrative subunits (that is, regions). NUTS 3 are still smaller units within NUTS 2. NUTS codes follow an identifiable structure. For example, Spain (ES) has an Eastern region (ES5), within which is the Autonomous Community of Catalonia (ES51), home to the province of Girona (ES512). Thus, all subunits are entirely nested within higher units and the NUTS code of any administrative district includes its location in within all superordinate boundaries. 40. Eurostat, http://ec.europa.eu/eurostat/web/nuts/overview (accessed 15 March 2015). 41. Arjan Schakel and Emanuele Massetti, “Ideology Matters: Why Decentralization Has a Differentiated Effect on Regionalist Parties’ Fortunes in Western Democracies,” European Journal of Political Research 52(6): 797–821 (2013). 42. Dan Miodownik, The Emergence of Demands for Regional Autonomy: Computer Simulation and European Evaluation (doctoral dissertation, University of Pennsylvania, 2005). 43. As a robustness check, we also run the analysis with regional party as a dummy variable; the results are reported in Table 3. 44. In regions with more than one qualifying regional party, the electorally most successful party is listed. 45. Fearon and van Houten, “The Politicization of Culture,” 16–17; David D. Laitin, “What Is a Language Community,” American Journal of Political Science 44: 142–155 (January 2000). Official reporting of minority language usage varies widely across the 11 cases in this study. Some states (for example, Belgium, Spain, United Kingdom) make this information widely available; others (for example, France, Greece) for political and other reasons, do not collect, nor disseminate, official information on the use of minority and regional languages. A third group of countries acknowledge the status of a limited number

Downloaded by [185.120.125.129] at 19:40 04 May 2016

Determinants of Regional Political Distinctiveness

143

of languages (for example, German, Friulian, Ladin, Slovenian in Italy, Danish in Germany) while ignoring other languages (for example, other local languages in Italy, Low Saxon in Germany). But even where data are made available by nonofficial sources, its scope and the quality of the information varies significantly from place to place (for example, Meic Stephens, Linguistic Minorities in Western Europe [Wales: Gomer Press, 1976]); Euromosaic, http://www.uoc.edu/euromosaic/ (accessed 15 March 2015); Ethnologue, http://www.ethnologue.com (accessed 15 March 2015); Eurolang project, http://www.eurolang.net/ (accessed 15 March 2015); Minorities in Europe, http://www.minority2000.net/ (accessed 15 March 2015); Geonative project, http://www.geocities.com/Athens/9479/ (accessed 15 March 2015); UNESCO Red Book, http://www.helsinki.fi/˜tasalmin/europe_index.html#country (accessed 15 March 2015). As a result, the consistency and quality of historic language usage is highly variable across the 11 cases. 46. M. Paul Lewis, ed., Ethnologue: Languages of the World, 16th ed. (Dallas: SIL International, 2009). 47. For Belgium, French was selected as the state language given its historic prominence. 48. Laitin, “What is a Language Community?” 148. A full schematic for all languages included in this study is provided in Appendix Figure 1. The end of each branch represents a language in the dataset with state languages indicated with a border around the bolded, underlined name. Only languages occurring in the dataset are included in the schematic. 49. Dialects of state languages represent nodes beyond the address of the state language and, as a result, will still be calculated as 0, given that the number of nodes the dialect and the state language share is the same as the length of the state language. The subtraction of the calculation from 1 rather than simply using the calculation has no statistical impact. However, by transforming the data this way, higher values reflect greater distinctiveness, consistent with the other measures and simplifying the interpretation of the signs generated by the model. 50. To assess the potential skewing of the analysis by the inclusion of Basque regions, we also run the models omitting Basque regions from the analysis, indicated in Table 3 with a +. 51. James L. Arbuckle, AMOS (Version 23.0) [Computer Program] (Chicago: IBM SPSS, 2014). 52. The one-sided arrows provide standardize coefficients (that is, the amount of change in the dependent variable attributed to a change in one standard deviation unit of the independent variable) linking each of the predictor and the outcome variables; the numbers on top of two-sided arrows indicate correlations between the independent predictors; the R2 value for each of the two dependent variables appears above the its rectangle on the diagram; lastly, the effect of other unobserved indicators on each outcome variable is indicated by the standardize coefficients above associated with the arrows linking the ecliptic boxes labeled “Error.” The software also provides by default tabular outputs similar to those provided by most statistical packages, helping to assist the overall quality of the model. These include chi-square test and other statistics of the model fit, as well as probability values for each of the predictors and the error terms. 53. Due to the limited degrees of freedom, we were not able to simultaneously include direct effects for economic distinctiveness and distance; however, as discussed above, those direct effects were not significant and are, therefore, excluded from the analysis and diagram. Figures 2 and 3 in the Appendix present models that estimate the independent effect of the economy (B) and distance (C) on PDI. As indicated in the model neither economic distinctiveness nor distance from the center independently increase political distinctiveness. 54. The chi-square is a test of the absolute fit of the model. Its small value (.668) with two degrees of freedom, and the associated probability (p = .716), suggest that the null hypothesis that the model fits the data well can be accepted with confidence. The Tucker-Lewis Index (TLI) compares the absolute fit of our model to an independence model, a more restrictive model that contains estimates of the observed variables only, in other words, and the Root Mean Square Error of Approximation (RMESA) provide further indication of the fit of our model. It assumes that the relationship between the observed variables is 0; the TLI value of 1.049 and RMSEA of .000 evidence the goodness of fit of our model (compare to Litze Hu and Peter M. Bentler, “Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives,” Structural Equation Modeling: A Multidisciplinary Journal 6(1): 1–55 (1999): rule of thumb TLI>.95 and RMSEA