Territory, place, scale, and polarization of electoral preferences in Ukraine, 2002-2014

Introduction

Not all newly emerged as well as established democracies have patterns of territorial homogenization of voting behavior in the process of nationalization of electorates and party systems. Communication technologies were gradually creating opportunities to reach the whole national population and preconditions for the diminishing of spatial differences in political beliefs. However, as overwhelming geographical research shows, regional and local contexts, geographical distance and existing and phantom borders are still essential for the understanding of elections outcomes. Moreover, social media can cause polarization of political beliefs (Bail et al., 2018; Bernhardt, Krasa, & Polborn, 2008; Prior, 2013; Tucker et al., 2018). It can be manifested in electoral preferences and may have territorial dimension. Moreover, Wing and Walker (2010) argue that geography is central to electoral polarization in the United States, Johnston and Others (2004) and McAllister and Studlar (1992) reveal regional polarization of electoral preferences in the United Kingdom, and Ferreira Do Vale (2015) shows that Brazil was also polarized along regional divisions in 2014 elections. Therefore, in contrast to the nationalization of party systems and territorial homogenization hypotheses, territorial polarization is still in question for new and established democracies.

Polarization is also about cleavages, and each party that wants to remain at political arena should crystalize cleavage in society to mobilize electorate along its lines (Lipset & Rokkan, 1967). Cleavages (as well as polarization) could follow territorial or functional lines in Lipset and Rokkan model; however, they are not strictly located in one pole of the divide. While cleavages along the territorial axis are geographical in their nature and, in Lipset and Rokkan theory, are about center-periphery relations, cleavages along the functional line (ideology or economic interests) may also have geographical patterns of manifestation in the political system (for example, the geography of uneven development). Maggini and Emanuele (2015) study of the nationalization of electorates and party systems in Europe reveals that in nationalized contexts the impact of the right- left party support is higher than in territorialized contexts. In cleavage theory, it could be interpreted as the nationalization of party system results in cleavages and polarization turn from the territorial to functional axis. However, conflict nature of politics and the existence of division lines remain essential for political mobilization.

Political beliefs and electoral preferences in Ukraine are broadly interpreted in terms of regional divisions (Barrington & Herron, 2004; Birch, 2000; Craumer & Clem, 1999; Katchanovski, 2006, 2014; Khmelko, Semenova, Teleshun, & Titarenko, 2011; O'Loughlin, 2001; O'Loughlin, Toal, & Kolosov, 2016), which also often include mentions of polarization (D'anieri, 2007; Kubicek, 2000; Kudelia, 2014; Sasse & Lackner, 2018; Braychevsky, 2012; Shyshatskyi, 2006) and cleavages (Barrington, 1997, 2002; Dawson, 1997; Katchanovski, 2006; Malanchuk, 2005; Zimmerman, 1998). These studies often are limited to the artificial construction of regions and compositional effects, which do not reveal the whole spatiality of political beliefs in Ukraine. The issue of the territorial homogenization of electorate in Ukraine during the independence years is not revealed. The opposite process, territorial polarization, should also be examined in more details and its dynamics. Despite focus on such narrow issue as territorial polarization and homogenization, the paper also accentuates that geographical theory developments, especially debates on the scale, politics of scale, place, territory, and networks use in geographical studies, have a great potential to be used for the better understanding of the complexity of the spatiality of political beliefs in Ukraine.

Literature review

In this section of the paper, we want to discuss recent debates on the scale, territory, and place concepts use in geographical research and their potential use in the interpretation of elections results and territorial polarization. Agnew (1997; 2002; Shin & Agnew, 2008) case study of Italy reveals how political parties use different geographical scales in their rhetoric, divide and structure territory of the country, and mobilize people in different local contexts. He assumes that geography is not external to political parties but inscribed in party politics and organization (Agnew, 1997). Agnew introduces politics of scale concept to electoral geography studies and argues that political parties in their rhetoric and organization are the main actors in the politics of scale. However, Brenner (2001, p. 602) is critical of such use of the scale concept stating that Agnew's case study which demonstrates scalar effects could not be narrowed only to scale and reveals complex spatialities of political processes, politics of location, and politics of territorial identity. As a result of the continuing debate on scale concept use in geography, Jessop, Brenner, and Jones, (2008) propose the TPSN framework (Territory, Place, Scale, Network). They argue that sociospatial relations should be understood as mutually constitutive and relationally intertwined dimensions of territories, places, scales, and networks because any singular of that dimensions cannot grasp the whole complexity of sociospatial processes (Jessop et al., 2008). Each of these four dimensions has its principle of sociospatial structuration: territory - bordering and enclosure, place - proximity, areal differentiation, and spatial embedding, scale - hierarchization and vertical differentiation, networks - interconnectivity and interdependence (Jessop et al., 2008, p. 393). The example of how place and scale are intertwined and mutually constitutive in electoral geography studies could be found in Johnston and his Colleagues (2005) study of neighborhood effect in Great Britain. They state that neighborhood effect operates at two levels: (1) level of social interactions in a locale, (2) level of a broader context, which also includes effects of local media and common agenda (Johnston et al., 2005). Agnew and Shin (2008) also show how messages at national level receive different local responses, which could be a result of the mediation role of party organization or place-specific factors.

Ukrainian electoral geography research is mainly limited to the regional level and does not problematize local contexts or scalar effects. We argue that scale and place are also essential for the better interpretation of electoral behavior in Ukraine. Swyngedouw (2001) in his political economy studies assumes that "place matters, but scale decides", which also could be applied to electoral geography studies. MacKinnon (2011) pointed out (proposing to replace "politics of scale" concept with "scalar politics") that different political actors can deploy scale strategically. Parties are such actors as was revealed by Agnew (1997; 2002). Moreover, geographical generalizations can influence political parties decisions how and where to campaign (in Johnston, Shelley, & Taylor, 1990, p. 228). For instance, the vision of one of the scales as essential for the electoral campaign success by political leaders or their consultants can shape the electoral geography of the country. At the same time, parties have limited resources for the organization and how they deploy them have a geographical component. There is also a problem understanding scale just as a level of analysis. For instance, Barrington and Herron (2004) applied two, four, and eight regions model for the analysis of compositional characteristics impact on electoral preferences in Ukraine and found that language was a statistically significant factor only in the case of 2 regions. In such and other cases, scale effect could be understood in Smith (1992) terms as a mediation between homogeneity and difference.

Data and Methodology

The paper is based on cartographical analysis of electoral outcomes at polling stations level at 2002-2014 parliamentary elections (only proportional vote in case of mixed electoral systems in 2002, 2012, and 2014). The period of analysis is limited to 2002-2014 because there is no data at polling stations level for parliamentary elections of 1994 and 1998 in Ukraine, which complicates the analysis of electoral preferences dynamics at the local level. To analyze territorial homogenization, we followed Caramani (2002) methodological remarks and applied the analysis of variance using mean absolute deviation divided by two to avoid double counting also known as Lee index (Lee, 1988). It means that we took the sum of each party share deviation from the national average and divided it by two for each polling station in Ukraine. Then we calculated average Lee indexes for each parliamentary elections in 2002-2014 for Ukraine and separately for oblasts to reveal regional variations in the dynamics of territorial homogenization. Data for 2014 parliamentary elections is without polling stations in annexed by Russia Crimea and uncontrolled by Ukrainian government territories in Lugansk and Donetsk oblasts. We also mapped Lee indexes and analyzed the territorial distribution of polling stations with the highest deviations from the national average.

Results and discussion

Lee indexes for parliamentary elections 2002-2014 (divided by 100 for calculations purposes) calculated at polling stations level reveal gradual territorial homogenization of Ukraine from 2002 (0.42) to 2014 (0.28) (see Table 1). At the same time, the country was facing territorial polarization because despite the decrease in national average mean absolute deviations, three Galician oblasts had a permanently high level of polarization in 2002-2012, and Donbas oblasts also had high Lee indexes in 2002-2014. Regional dynamics in comparison to national shows that Galicia was polarized in 2002 in contrast to other regions of Ukraine, while 20062012 period was characterized by two poles territorial polarization, while at the national level territorial homogenization increased. In 2014 parliamentary elections, Galician oblasts homogenized to national mean, while Donbas oblasts remained polarized.

Table 1

Mean absolute deviation divided by two for Ukraine and Donetsk and Lviv oblasts for parliamentary elections 2002-2014 calculated at polling stations level