CEPSA 2011, Vienna - Multilevel Politics: Intra- and Inter-Level Comparative Perspectives

CEPSA 2011, Vienna - Multilevel Politics: Intra- and Inter-level Comparative Perspectives

West and East: Still Mind the Gap?

Aggregate Analysis of Electoral Behavior in Central Europe[1]

Petr Voda – Michal Pink

First draft only for presentation at conference venue

Abstract

A series of changes has taken place in Central Europe after the fall of communism. Some of these changes have been driven by the increased influence and relevance of a multilevel government, the highest level in the hierarchy being Brussels and the lowest, the closest municipality. A question remains as to whether these changes have been of a uniform nature, or whether they have influenced different states – and even different locations within the same state – in different ways. This paper focuses on patterns of electoral behavior in terms of the spatial distribution of election results and the impact of social structure. It asks whether these patterns have become more similar in the Central European region. The basic hypothesis is that the influence of socioeconomic characteristics is becoming ever more similar with the continuing integration of Central Europe. The first phase of the analysis will be quantitative in nature. We employ ordinary least squares regression and geographical weighted regression, using data from elections and censuses, to describe similarities and differences in the influence of electoral behavior determinants among countries (Austria, the Czech Republic, Slovakia and Hungary), as well as the development of these factors. The second phase studies the alignment of results in terms of important dates in the integration process.

Introduction

The countries of the EU have experienced different histories over the last half-century. A key distinction dividing members is between those countries in the East which were under nondemocratic communist regimes and those of the West, where development was democratic. Countries on both sides of this division lie side-by-side in Central Europe, with the dividing line running through Germany, between Germany and the Czech Republic, Austria and the Czech Republic and Austria and Slovakia, Hungary and Slovenia. At times in the past, these countries have been very close – almost identical in many respects – as well as very different, as was the case from the 1950s until the late 1980s. Since that time, they have once again grown closer and likely more similar, as well.

Over the last 20 years, many new actors have appeared on the political stage in these countries, particularly on the eastern side of the dividing line. Numerous integration projects have taken place in the region. The most influential would be the coming of the EU. Austria acceded in 1995. Poland, the Czech Republic, Slovakia, Slovenia and Hungary became part of the EU in 2004, after a lengthy entrance procedure. These milestones have played an important role in changing the politics, policies and polities of countries in the region. European integration are obviously very important for changing politics, policies as well as polities of all these countries (see Carbone 2010, Goetz, Hix 2001, Krieger-Boden, Morgenroth, Petrakos 2008 etc.). But how deep are these changes? A lot of articles suppose some change in manifestos of parties. However, we do not know, how affected voters are.

Our focus will not be on parties, policies or actors, but rather on the relationship between parties and their voters in elections. We make the following assumptions: just after the revolutions of 1989, Western and Eastern countries differed greatly from each other. In the West, political parties obtained votes from different social groups and made different policies. Over time, though, with capitalism taking hold, increasing wages and unemployment and, of course, with the EU integration process, both societies and politics in the East have changed.

The question is: how have connections between parties and society developed in Western and Central Europe?

The answer is not easy to find. Data from several sources and multivariate statistical methods will be used to provide one. Manifestoes, party activities, governments and parliamentary bodies will not be the focus; we will instead concentrate exclusively on the election results for relevant parties in general elections. There are several potential ways to work with these results. We will use data for the micro-regional level and employ "traditional" OLS regression. To address problems associated with OLS regression in studying nonstationary relationships, geographically weighted regression techniques will be used to take special notice of the manner in which the relationship between election results and certain causal factors varies across space (see Kavanagh 2006).

Regression analysis will be used to make comparisons. The focus will be on determining whether parties within the same family but in differing countries show similar support patterns and whether the changes these patterns undergo are similar in nature. It is here that a key issue arises. In progressing from counting GWR and interpreting the results to making comparisons between countries, we suddenly have only a highly restricted space to control for. Thus it is not certain whether the changes observed have truly arisen from the causes theorized, or whether another factor may be impacting the development of electoral behavior in the countries selected.

To determine an answer to the second question, Rokkan’s theory of cleavages will be applied. There are for original cleavages: urban vs. rural, center vs. periphery, owners vs. employees and church vs. state (Lipset, Rokkan 1967: 9-23). Each of these cleavages could give rise to a party. The four cleavages are based upon the structure of society during the 19th century, when parties arose around conflicts between the aims of the urban and rural population, between owners and employees, between religious and secular individuals and between inhabitants of the center and the periphery. Particular sides in the conflict have changed over the course of the last century, but its bases likely remain the same. Some people are more motivated to vote for certain parties as opposed to others, because their social status differs (Evans 2004:42-68).

For finding an answer to the question Rokkan’s theory of cleavages will be applied. There are four original cleavages: urban – rural, center – periphery, owners – employees and church – state (Lipset, Rokkan 1967: 9-23). On each of these cleavages some party could exist. These four cleavages are based on the structure of society in 19th century, when parties arose around conflicts between aims of urban a rural population, between owners and employees, between religious and secular and between inhabitants of the center and periphery. These sides of conflict changed throughout the last century but the bases are probably still the same. Some people are more motivated to vote certain parties than others because they have a different social status (Evans 2004: 42-68).

Several articles dealing with electoral behavior in early 90’s in post-communist countries assume the Rokkan’s theory inappropriate in context of new democracies because of historical and geographical bases of this theory as well as specific context of states after falling communist regimes. Dalton (2000: 925-926) notices that emerging party systems are unlikely to be based on stable group-based cleavages, especially when the democratic transition happened rapidly as in Eastern Europe. Also new electorates are unlikely to hold long-term party attachments that might guide their behavior. That is the reason, why the patterns of electoral choice in new democracies are more involved by short-term factors like candidate images and issue positions.

Several studies also exist concerning the impact of European integration on electoral behavior. Gabel (2001: 52-54) sees research on the topic focused above all on the question of issue voting (see de Vries 2007). He notes the example of Great Britain, in which EU issues have created new electoral cleavages, and examines the importance of European monetary policy in economics-based voting behavior. Tillman (2008) also addresses the impact of attitudes towards European economic issues on voting behavior and national politics.

Such an approach imposes a number of limitations, especially in the comparative dimension of analysis. The data itself may be hard, but changing conditions and modified meanings over time and through space make it function rather more like soft data. The fact that a vote was cast for certain party in certain elections in certain country does not entitle us to compare the meaning of the vote with one cast for similar party in election in another country. The meaning may be better described by applying more general concepts. Thus the Rokkan’s theory and Bayme’s classification of parties are used. But this will only mitigate, not eliminate the limitations. These concepts may help us to locate cases in which the level of comparability is higher than in others.

Other way to provide comparability is to select countries on an areal basis. We chose Austria, the Czech Republic, Slovakia and Hungary with logic of Mill’s method of most similar cases. All countries have similar history and cultural bases. The countries share common frontiers, allowing the degree to which societies are contained within the geographic area of the country to be determined, as well as what development is like in the Western and Eastern blocs. These countries are also similar in terms of their political regimes. Each country has a parliamentary democracy. The only distinction lies in the unitary character of the Czech Republic, Slovakia and Hungary; Austria, by contrast, may be described as a federal state. All states, however, have autonomous regional governmental elections, meaning the parliamentary elections have almost the same meaning and sense in all of these countries.

The party systems of all the countries under consideration are fairly close. All of them are pluralistic as regards the number of parties and are moderate as regards type under Sartori's classification. The only exception is the Czech system, which is assumed to be semi-polarized (Strmiska 2005). However, individual systems differ in time and space in terms of the number of relevant parties. This probably forms the fundamental limitation on our conclusions and the research as a whole. This becomes even more critical when we focus on development. It is obvious that the party system has changed markedly since 1990, but it may be less obvious that the parties themselves have also changed. There is no party in which the same personalities are saying the same things as was said in 1990. Questions related to party systems are important because the nature of a party’s system creates possibilities for voters. From our perspective, the different structure of party systems might result in different bases of electoral support for the parties.

There is a real threat that the composition of a party's electoral base will not adequately represent the party's ideological profile. This depends upon the voters' ability to make decisions concerning their vote. Their votes may be influenced by several other factors described by psychological models and by issue-based models, for example: sympathy for a particular candidate, identification with a party, or agreement with the party's attitude on some key issue. All of these factors taken together cannot be captured in a single article. In this case, then, the research problem has been simplified to include only the influence of social stratification on election results.

Several methodological issues arise in terms of the comparability of electoral systems as well as party systems. Chief among these are the Hungarian elections, which take place under the rules of a mixed system (see Sedo 2009). Collections in the other countries (Austria, the Czech Republic and Slovakia) use a proportional system. The problem arises because no election results are available for areas smaller than electoral districts. Thus, results from the majority party must be used. This represents a huge limitation in making comparisons, since the conditions of the majority system are quite different to those in the proportional system. In particular, there are strong incentives for strategic voting. Smaller parties are heavily impacted, because there is almost no point in voting for them in a majority system (see Abramson, et al. 2010). Hungary will nevertheless remain in the analysis but these potential limitations must be borne in mind.

Data

The research question makes clear that data is needed concerning elections and societal factors. The data used in this analysis derive from two sources. Electoral results for each general election are taken from the central election commissions of the countries selected. These election results will serve as the dependent variable for the first portion of the study. The dates at which the elections were held are suitable for purposes of comparison. Every country involved had elections in 1990, 2002 and 2006. There were elections in three countries in 1994, 1998 and 2010 (see Table 1). Temporal context, then, presents virtually no problem.

Table 1: General Elections in Central Europe in 1990 - 2010

Source: Parties and Elections

Societal data is derived from censuses that took place in 1991 and 2001 in all selected countries. Independent variables have been selected using a multi-step process. First, cleavage indicators under Rokkan’s theory are selected. All independent variables have some relationship to a particular cleavage. The following variables have been selected: proportion of self-employed persons, proportion of highly educated persons and the unemployment rate as indicators for the economic cleavage; proportion of agricultural workers and urbanization rate as indicators for the urban-rural cleavage; ethnic divisions in society for the center-periphery cleavage; and for the church-state cleavage, the proportion of Christians.

In the second step, variables with no comparable data and variables lacking data for all countries and censuses are eliminated. This leaves only the proportion of Christians, agricultural workers, persons with a university education, unemployment and the rate of urbanization. Finally, all variables demonstrating high global or local multicollinearity are eliminated. The resulting list of variables is as follows: proportion of Christians, proportion of agricultural workers and number of highly educated persons.

All data are tied to regional NUTS4 units, except for data relating to Hungary. The units concerned are called ORP in the Czech Republic (Municipalities with Extended Powers), Okres in Slovakia and Politische Bezirke in Austria. Results for Hungarian general elections were available only for electoral districts, while census data was available for different census regions. Data for the analysis was calculated using overlapping areas. Table 2 shows the basic characteristics of regions. The number of units is also number of cases of further regression analyses. Austrian, Hungarian and Slovakian units are very similar in average number of inhabitants as well as average area. Czech micro-regions are smaller.

Table 2: Regional Units of Analysis

Source: own calculation, based on Statistik Austria, ČSÚ, SŠÚ and valastazs.hu

Coefficient raster workspace (optional)

A full pathname to the workspace where all of the coefficient rasters will be created. When this workspace is provided, rasters are created for the intercept and every explanatory variable.

Output cell size (optional)

The cell size (a number) or reference to the cell size (a pathname to a raster dataset) to use when creating the coefficient rasters.

The default cell size is the shortest of the width or height of the extent specified in the Environment output coordinate system, divided by 250.

Prediction locations (optional)

A feature class containing features representing locations where estimates should be computed. Each feature in this dataset should contain values for all of the explanatory variables specified; the dependent variable for these features will be estimated using the model calibrated for the input feature class data.

Prediction explanatory variable(s) (optional)

A list of fields representing explanatory variables in the Prediction Locations feature class. These field names should be provided in the same order (a one-to-one correspondance) as those listed for the input feature class Explanatory Variables parameter. If no prediction explanatory variables are given, the output prediction feature class will only contain computed coefficient values for each prediction location.

Output prediction feature class (optional)