Economic and political effects on European Parliamentary electoral turnout in post-communist Europe

Electoral Studies 27 (2008) 661–672

Contents lists available at ScienceDirect

Electoral Studies journal homepage: www.elsevier.com/locate/electstud

Economic and political effects on European Parliamentary electoral turnout in post-communist Europe Christine Fauvelle-Aymar a, Mary Stegmaier b, * a b

ˆpital, 75647 Paris Cedex 13, France Maison des Sciences Economiques, LAEP, University of Paris 1, 106–112 bd de l’Ho College of Arts and Sciences, University of Virginia, P.O. Box 400133, 101 Garrett Hall, McCormick Road, Charlottesville, VA 22904-4133, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 16 April 2007 Revised 27 February 2008 Accepted 29 May 2008

The relatively low voter turnout rates in the June 2004 European Parliamentary elections in many of the post-communist states surprised observers. While the average turnout rate for these new-EU member states barely surpassed 30%, turnout exhibited much variance at the national and sub-national levels. In this article, we study the economic and political determinants of European Parliamentary voter turnout in the post-communist countries using a unique region-level dataset. Our regression results reveal that regional unemployment rates have a statistically significant impact on turnout. Regions with higher unemployment rates experienced lower turnout, even after controlling for political and socio-demographic factors. In contrast to some previous work on the impact of EU support on EP turnout, our study uncovers a positive relationship between these two variables. Further, we show that the timing of the election relative to the next national election and the frequency of elections affected turnout. Ó 2008 Elsevier Ltd. All rights reserved.

Keywords: European Parliament Elections Turnout Economics Eastern Europe European Union Post-communist elections Political participation

1. Introduction On May 1, 2004, with much fanfare, the European Union extended membership to the first group of post-communist countries: Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia and Slovenia. After the fairly high voter turnout in the referenda, the sizeable ‘‘yes’’ vote in these polls, and much political hype about these countries’ ‘‘return to Europe,’’ many observers were surprised by the paltry voter turnout in many of these countries during the June 2004 European Parliamentary election. While turnout rates varied across these countries, the overall average barely surpassed the 30% mark. Fig. 1 illustrates the turnout rates in the European Parliamentary (EP) election and, as a reference point, the European Union (EU) referendum in these eight countries.

* Corresponding author. Tel.: þ1 434 924 8864. E-mail addresses: [email protected] (C. [email protected] (M. Stegmaier).

Fauvelle-Aymar),

0261-3794/$ – see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.electstud.2008.05.008

In each case, turnout in the EP election was markedly lower than in the referendum, with average turnout as a percent of registered voters at 30.58% and 58.46% respectively. The range in EP turnout rates is quite remarkable. Lithuania led in voter turnout with 47.97%, while in Slovakia only 16.97% of registered voters cast a ballot. At the regional level, a Polish region posted the lowest turnout, 14.16%, and a Lithuanian region exhibited the highest turnout, 52.54%. Furthermore, the average turnout rate in these countries was also lower than the average for the 15 established member states, where analysts have been monitoring the steady decline of turnout rates over the past decades (Adshead and Hill, 2005; Delwit, 2002; van der Eijk and Oppenhuis, 1990).1 The post-communist new member states offer a unique laboratory to test explanations of aggregate voter turnout. These countries have had shared experiences with

1 The average turnout rate for the 15 countries was 47.79% of registered voters.

662

C. Fauvelle-Aymar, M. Stegmaier / Electoral Studies 27 (2008) 661–672

80

Referendum European elections

70

2.1. Economic conditions

60 50 40 30 20 10 0

.

ep

R ch

ze

C

2. Literature and hypotheses

ia

ton

Es

y

ar

ng

Hu

a

ia

tvi

La

an

hu

Lit

d

lan

Po

ia

ak

ov

Sl

ia

en

ov

Sl

Fig. 1. Turnout rates in the 2003 EU referenda and 2004 EP election.

communism and the economic and social upheaval caused by the economic transformation process in the 1990s (Blazyca, 2003; Cox, 2003; Lavigne, 1999; Roland, 2000; Vecˇernı´k and Mateˇju˚, 1999). The EU accession negotiations prompted further economic reforms in order for these countries to comply with EU standards (Landesmann and Rosati, 2004; Nugent, 2004). The speed and impact of the economic reforms have varied across these countries (Vachudova, 2005), but their recent economic experiences set them apart from the established EU member states and from Malta and Cyprus, which also joined the EU on May 1, 2004. When we compare the per capita GDPs of the established EU member states and the post-communist new members, we find a clear division in their economic standards. Each of the post-communist countries posts a per capita GDP below the average of the established member states. Further, at least during the initial years of EU membership, all are net beneficiaries (McCormick, 2005; Nugent, 2004). While these countries share similar economic circumstances which are distinct from the West European experience, there are also some political commonalities for the purposes of studying aggregate turnout. The post-communist countries all voted in referenda on EU accession during 2003, which provides us with a timely and practical measure of EU support across the countries. Additionally, many of the electoral features of EP elections that have explained variance in national turnout rates in West European EP elections do not vary across the post-communist countries. Given the extent of shared commonalities, what factors explain this high variance in aggregate turnout rates among these post-communist new-EU member states? In this article we address this question and argue that regional economic conditions play a central role in explaining turnout at the regional level. Since economic outcomes have varied greatly within countries during the transition, regional economic winners and losers have emerged with more or less at stake in the EP elections. Further, we expect that political measures, such as support for the EU, timing of the EP election, and election frequency will contribute to the explanation. It is to these hypotheses that we now turn our attention.

As noted above, these eight post-communist countries share common economic experiences; however, the impact of the economic reforms has been uneven within the countries. In the East European capitals and large cities that have transformed into centers of business, tourism or modern industry, the economies have prospered. People in these areas have job opportunities in fields that often would benefit from greater integration with the rest of Europe. In rural areas and regions reliant on outdated heavy industry, unemployment tends to be high, standards of living are low, and international opportunities that expose citizens to the benefits of joining Europe are few and far between. Since stark economic differences have emerged within these countries, we could expect to find significant differences in political behavior. First, one can expect the economic situation to have implications on the citizens’ opinions about the European Union. Residents of prosperous regions may in general be more likely to support the EU, because they see the impact of internationalization on their own economic well-being and opportunities. Conversely, for those in economically stagnant or depressed regions, the economic outlook may be bleak. They see people around them struggling to find work or to find more meaningful work that utilizes their skills, and they are likely to perceive that they are the losers in economic transformation. They may oppose the EU because of the type of economic system it represents, or because they see no potential benefits for their region from membership. This argument has found tentative support in research by Christin (2005), Tucker et al. (2002), and Tverdova and Anderson (2004). Another area of political activity where these differences have already been demonstrated is election outcomes. Various studies of post-communist countries have shown that the economic position of voters shapes their support for pro-reform versus anti-reform parties, or using the terms of Tucker (2006), Old Regime versus New Regime parties. Studies conducted at the province- or districtlevels in post-communist countries have demonstrated that areas with poor economic conditions, especially high unemployment, are more likely to support parties that are opposed to further liberalizing economic reforms (Bell, 1997; Fidrmuc, 2000; Tucker, 2006; for a review, see Lewis-Beck and Stegmaier, in press). Since economic conditions have affected support for the EU and election outcomes, it seems that this could be an important factor in explaining turnout. Economic factors are not commonly included in studies of voter turnout and electoral participation, though the theoretical argument and evidence that economics matters has long existed. Rosenstone (1982), in his seminal work on economic adversity and voter turnout, reasoned that economic duress would cause voters to withdraw from political participation. Instead of following an election, they would need to devote their time and energy to meeting their basic needs. These citizens would be searching for

C. Fauvelle-Aymar, M. Stegmaier / Electoral Studies 27 (2008) 661–672

663

2.2. EU support

should be more likely to participate in electing representatives to its governing body. While this expected relationship intuitively makes sense, research on voter turnout in the older EU member countries has provided spotty evidence for the hypothesis. A few recent studies of EP turnout have found that turnout is higher in countries where support for the EU is greater (Flickinger and Studlar, 2007; Mattila, 2003), but a number of other aggregate analyses have failed to find evidence that EU perceptions matter for turnout (Franklin et al., 1996; Rose, 2004; Schmitt, 2005; Schmitt and Mannheimer, 1991). For example, in his study of turnout in all member states for the 2004 election, Rose (2004) uncovered no statistically significant relationship in bivariate correlations between the percent in the country who believe that their country will benefit from EU membership and turnout. Blondel et al. (1997) have openly questioned these findings by asking: ‘‘Can it really be that attitudes to Europewhat people know and think and feel. about the European Union and its institutionsdplay no role in determining whether or not they vote in a European Parliament election?’’ (p. 246). Their individual-level study of the 1994 EP election focuses on reasons for why abstainers abstained from voting. Some abstained for circumstantial reasons, but more interesting for our purposes are the reasons offered by the voluntary abstainers. The most prominent reason given by the non-voters was lack of interest in the election. Other explanations included distrust of politics, lack of knowledge, and the belief that one’s vote has no consequence. Opposition to the EU was a reason cited by 8% of non-voters in the EU countries with non-compulsory voting. However, of importance for the study of post-communist turnout in EP elections, Blondel et al. provide the distributions for East and West Germany separately. In East Germany, opposition to the EU is cited by 22% of non-voters as the reason for not votingdthe highest percentage of the 10 countries listed. West Germany trailed in a distant second place with 13% reporting this reason for not voting (Blondel et al., 1997, p. 255). This could be because of the EU’s ‘‘newness’’ to the former East German citizens, or it may have to do with the communist legacy. Regardless, the finding suggests that support for or opposition to the EU may have a more substantial impact on turnout in the new member post-communist countries than in the established member states. Furthermore, there are additional reasons to believe that support will matter for turnout. Politicians and the media paid close attention to the pros and cons of EU membership during the run-up to the 2003 EU referenda, which enabled citizens to form opinions concerning membership. It is unlikely that many voters changed their minds about membership in the short time span between the referendum and the first EP election. While the referenda passed in all these countries, the results were far from unanimous.2 Of the citizens who opposed their

Researchers focusing on turnout in EP elections have posited that attitudes concerning the EU are important for participation and abstention. Citizens who are knowledgeable about the EU and hold favorable opinions of it

2 For EU referendum results, see the respective country national election committee websites. A convenient listing with links to these sites can be found at http://electionresources.org/eastern.europe.html

work, or working extra jobs, in order to feed and house themselves and their families. Rosenstone tested this argument using both aggregate- and individual-level data from the U.S. and found support for his hypothesis that economic hardship leads voters to withdraw from the political sphere (the withdrawal hypothesis). While he considered the counter-argument that adversity would drive voters to the polls to punish the incumbent (the mobilization hypothesis), he found no evidence that this occurs in the U.S. Scholars have tested this economic adversity argument in Eastern Europe (Pacek, 1994; Tworzecki, 2003) as well as in other political contexts (Lewis-Beck and Lockerbie, 1989; Radcliff, 1992). Pacek (1994) analyzed district-level data from three Central European countries and found support for the withdrawal hypothesis. After controlling for turnout in prior elections and education levels, the results revealed the expected negative relationship between the unemployment and turnout. At the individual level, Tworzecki (2003, p. 168–9) found that the unemployed in Poland and the Czech Republic were less likely to turn out than those who were working. Applying the logic of the withdrawal hypothesis, we can expect that a higher portion of citizens who reside in economically weak regions of Central and Eastern Europe will be focusing their attention on meeting their basic needs rather than informing themselves about politics, particularly politics at a supra-national level. Further, while it takes time to learn about an election, it also takes time to go to the polls to vote. Everyone has to prioritize the use of their time, and thus, it’s reasonable to expect that when a person is unemployed, he or she may rank voting below tasks that could improve the person’s economic well-being. There is another possible reason to believe that people in these economically stagnant regions would refrain from participation. If citizens feel that their region has been hurt by the increased economic competition, privatization, the removal of state subsidies, and other reforms implemented in the economic transition, they might be ambivalent toward the EU. If people do not perceive that the EU will impact their community, which is more likely to be the case in depressed or rural areas, then we can expect to see lower turnout. Regardless of whether the economic problems lead voters to refrain from participation because they have other more pressing needs, or because they don’t think the EU matters to them, the hypothesized relationship between regions with relatively weak economies and turnout is the same. H1: A strong regional economy should lead to higher regional turnout rates and a weak regional economy should produce lower regional turnout rates.

664

C. Fauvelle-Aymar, M. Stegmaier / Electoral Studies 27 (2008) 661–672

country joining the EU, many likely viewed the EP election as irrelevant and thus abstained from voting. H2: Higher levels of EU support should lead to higher turnout rates, while lower levels of EU support should lead to lower turnout rates.

2.3. Political factors The vast literature on voter turnout in advanced industrial democracies has provided the foundation for constructing models of EP election turnout (for recent thorough treatments of this literature, see Aarts and Wessels, 2005; Blais, 2000; Franklin, 2002, 2004; Geys, 2006). While the secondorder nature of EP elections differs from national-level elections,3 the primary factors affecting turnout rates are quite similar. At the aggregate-level, scholars have followed the lead of Glass et al. (1984), Jackman (1987), and Jackman and Miller (1995) by testing the impact of various electoral structures and rules across countries. In studies of EP turnout prior to the 2004 EU expansion, the variance in national-level turnout rates could be explained well by factors such as weekend voting, compulsory voting, concomitant elections, and the type of electoral system employed for the election (Franklin et al., 1996; Mattila, 2003). In their aggregate national-level study of turnout in all EU member states in the 2004 election, Rose (2004) and Flickinger and Studlar (2007) tested for a host of institutional, attitudinal, and EU related factors. While Rose (2004) did not test a multivariate model, the results are suggestive. Like Mattila (2003), he found the expected bivariate correlations between turnout and the structural factors of compulsory voting and concurrent national elections. In 2004, however, all countries used a form of PR, so the electoral system did not affect turnout rates. Rose’s results also revealed that higher levels of trust in domestic political parties and the national government correlated with higher levels of turnout. The length of EU membership was also related to turnout, with the established EU countries showing higher turnout than newer member states. While these findings greatly contribute to our understanding of how electoral practices affect turnout rates, many of the explanations do not contribute to explaining turnout variance in this study. In all of the post-communist countries the PR electoral system is used for EP elections, all held their election on, or at least partially on, a Saturday or Sunday, and as a reaction to their communist pasts, none has compulsory voting. However, there are two features that do vary across these countries: timing of national elections relative to the EP election and the sheer number of elections held in the country’s recent past.4

3 Electoral scholars have long established that EP elections fit the second-order election conditions (Hix and Marsh, 2007; Marsh, 1998; Reif and Schmitt, 1980; Schmitt, 2005). A defining feature is that the electorate believes less is at stake than in a first-order national election and thus fewer citizens turn out to vote. 4 There are other institutional characteristics which differentiate these countries, such as their party system and features of their electoral campaign, but the two features we account for have the advantage of being easily quantifiable and then introduced in an empirical study.

First, as for the timing of national elections, there is one case of a concomitant national election, which also happens to be the post-communist country with the highest turnout in the EP election: Lithuania. In April 2004, the Lithuanian president, Rolandas Paksas, was impeached on charges of corruption, prompting an early presidential election that coincided with the EP election (Jurkynas, 2005). The events leading up to the presidential election captured the attention of Lithuanian voters, and, in turn, likely elevated turnout higher than it otherwise would have been. This leads to our first structural hypothesis: H3: In countries where national elections are concurrent with the EP election, turnout will be higher than in countries that do not have concomitant elections. If the concomitance of elections increases turnout, then more generally, the election calendar may also matter. According to the signaling dimension of voting, voters may use the present election as a method for sending a signal to the political market, both to other voters and to candidates (Fauvelle-Aymar et al., 2000). For example, voters may seek to influence a party’s program for the next election, and the vote is a means of doing this. As predicted by the second-order election model,5 voters will use lower salience elections, such as the EP elections, ‘‘as a low-cost opportunity to voice their dissatisfaction with government parties’’ (Schmitt, 2005, p. 652). While the second-order election model suggests that this will result in losses to the government, Franklin (2001) and others have argued that the timing of the EP election relative to the national election also has implications for turnout. When a general election is on the horizon rather than years away, voters have greater reason to expect that their signal will be heard and acted upon by the parties, since parties want to improve their chances of winning the general election (Fauvelle-Aymar et al., 2000). This increases the salience of the election to voters, which ought to produce higher turnout. Research on this by Franklin (2001) substantiates this hypothesis for EP elections from 1979 to 1999. This leads to our second structural hypothesis: H4: The shorter the time-span until the next general election (presidential or parliamentary) in the country, the higher turnout will be. The last political variable that we test for is the number of elections held in the country prior to the EP election. The past electoral calendar may influence turnout behavior since, in some cases, it may induce voter lassitude. For example, according to Auers (2005), turnout in the case of Slovakia might have been quite low due to voter fatigue, since the country had two rounds of presidential elections in the months prior to the EP election. If voters have

5 In addition to producing lower turnout, second-order elections typically generate losses for the governing and large political parties. Since voters perceive that less is at stake in these supra-national elections, voters use the election to signal their discontent with the national governing parties and to vote according to their ‘‘heart’’ rather than strategically for the major parties. Recent research on party vote-shares in the new member states has found that these aspects of the second-order election model hold less well than in the established member states (Hix and Marsh, 2007; Koepke and Ringe, 2006; Schmitt, 2005).

C. Fauvelle-Aymar, M. Stegmaier / Electoral Studies 27 (2008) 661–672

665

Fig. 2. The turnout rate in the 2004 European elections.

recently had the opportunity to voice their political views, especially in multiple elections, they may be less inclined to go to the polls again, especially when the election is a second-order election. This argument leads to our third structural hypothesis: H5: The higher the number of elections held in the recent past, the lower turnout in the EP election will be.

3. Data and methodology To study the determinants of regional voter turnout in the EP elections, we have compiled a dataset at a subnational level from the Eurostat REGIO database and country statistical offices. These data are disaggregated by region within each country at different levels referred to as ‘‘NUTS,’’ Nomenclature of Territorial Units for Statistics. The NUTS territorial units were established to provide standard regional statistics for the EU and are used for policymaking and analysis. The NUTS nomenclature is a threelevel hierarchical categorization that divides each of the 25

EU member states into a whole number of NUTS-1 regions. Each of these NUTS-1 regions is then divided into a whole number of NUTS-2 regions, and so forth.6 Despite the goal of creating comparably sized regions at the same NUTS level, each level still contains regions that differ in area and population. The NUTS-1 level in some cases, such as the Czech Republic, is the entire country. But in other countries with large populations, such as Poland, the NUTS-1 level encompasses multiple regions. In some countries, such as Estonia, Latvia, Lithuania, and Slovenia, because of their small populations, the NUTS-1 and NUTS-2 levels remain the entire country. However, at the NUTS-3 level, these particular countries are divided into multiple areas. We collected data on turnout rates, EU support, economic conditions and demographics at the NUTS-3 level, which, for our eight countries, produces a sample size of 119. The turnout and EU referendum data were collected from the National Electoral Institutes of each country. The economic and socio-demographic data came from the Eurostat REGIO database and when data were missing, we compiled them from the National Statistical Institutes of each country.7 Measures of election timing and frequency are national-level. While other scholars have

Table 1 2004 EP turnout in the post-communist countries

Czech Republic Estonia Latvia Lithuania Hungary Poland Slovenia Slovakia

Number of regions

Mean

Std. Dev.

Min.

Max.

14 5 5 10 20 45 12 8

28.01 25.49 40.56 47.94 37.07 20.31 28.33 16.97

3.08 2.97 2.70 2.15 4.51 4.72 2.76 1.46

23.11 23.35 38.30 45.17 30.46 14.16 24.30 15.49

34.61 30.38 43.83 52.54 49.56 35.51 32.39 20.24

6 More precisely, the NUTS nomenclature subdivides the territory of the European Union (25 countries) into 89 regions at NUTS-1 level, 254 regions at NUTS-2 level and 1214 regions at NUTS-3 level. More information on the NUTS nomenclature may be found on the Europa web site: http://europa.eu.int/comm/eurostat/ramon/nuts/home_regions_en.html. 7 Additionally, for some Polish and Latvian measures, NUTS-3 level data were unavailable. With the assistance of Eurostat, we were able to aggregate the data from smaller territorial units for these countries into the NUTS-3 units.

666

C. Fauvelle-Aymar, M. Stegmaier / Electoral Studies 27 (2008) 661–672

Fig. 3. The rate of unemployment.

employed regional or district data to study voting behavior in national elections in post-communist Europe, we are unaware of other research using the NUTS-3 territorial level to study voter turnout in cross-national studies. There are some key advantages to using NUTS-3 level data rather than national-level data. First, the larger sample size of 119 territorial units compared to just 8 countries affords us greater confidence in our statistical results. Second, we have much greater variance in many of our independent variables than we would at the national leveldparticularly with the education and age measures.8 This may result in different overall findings than we would achieve with national-level data. And third, striking disparities of income and opportunity exist within these countries that would be masked in a more aggregated analysis.9 At the individual level, these regional economic differences will likely lead to differences in political behavior and views of how the EU will benefit or harm their local interests. Nevertheless, since our analysis is at the aggregate level, even if the use of regional data offers the possibility to better assess the influence of the economic situation on turnout, the aggregated nature of the data does not allow us to make any inference about individual-level participation in order to avoid the ecological fallacy. In fact, our analysis is what is called ‘‘an ecological analysis’’ where the aim is to examine the relationship

8 For instance, the percentage of people with a post-secondary level of education (our education variable) varies between 9% and 56% when one considers the national averages, whereas it varies between 5% and 76% at the regional level. 9 For instance, the unemployment rate varies between 6.5% and 19.6% at the national level and between 3.8% and 30.9% at the regional level.

between the characteristics of the voters’ territories and the electoral results. How does the environment influence voters’ behavior? This is a very important question that has been largely neglected with the dominant use of individual survey data, where there is no indication of localization. The dependent variable in our models is the EP turnout rate in each NUTS-3 region. Turnout is calculated as the number of ballots cast divided by the number of registered voters. The variance in turnout is illustrated in Fig. 210 and Table 1. The economic measures used to test the impact of economic conditions on turnout are: the rate of unemployment in 2004 (shown in Fig. 3), the level of GDP per capita in euros in 2002, and the average growth rate of the GDP between 2001 and 2002.11 These measures are all calculated at the NUTS-3 level. Table 2 presents some summary statistics for our explanatory variables. Support for the European Union is measured as the percent of voters who cast a ‘‘yes’’ vote in the district during the European Union referendum in 2003. While many scholars who have incorporated a measure of EU support in their turnout models have used aggregated survey measures of support, we use the EU referendum results. While the referenda were not held in all eight countries on the same day, they were all held fairly close to one another (between March 2003 and September 2003)dnot too long

10 While the unemployment measures are available for 2004, national account data take longer to be published and thus the 2002 GDP data were the most recent data available. 11 Census data concerning education always refers to the population above 15 years of age, which is the working age population.

C. Fauvelle-Aymar, M. Stegmaier / Electoral Studies 27 (2008) 661–672

before the June 2004 EP election. Since the referenda results are available at the NUTS-3 level, it offers us the opportunity to take into account the variance of EU support within each country, which would not be possible to do with available survey data. One drawback to this measure is that it does not account for the opinions of those who abstained in the 2003 referenda. However, a similar bias exists with surveys, which do not account for the opinions of people unwilling to participate in the survey or who are unresponsive to the question. However, the enormous advantage of our indicator is its availability at the regional level. The average percentage of voters that cast a ‘‘yes’’ vote is 80.73% for these countries, but varies between a minimum of 48.43% in one Latvian district and a maximum of 96.09% in one Slovak district (as seen in Table 2). The regional variance can be seen in Fig. 4. We define two election variables. The first is the span of time (in days) until the next general election. This variable is set to zero in the Lithuanian case since the presidential election was held the same day as the EP elections. This means that we cannot test hypotheses H3 and H4 separately. The second election variable is the number of elections (except referenda) that were held during the 18 months prior to the EP elections (that is, since January 1, 2003). As controls, we introduce aggregate-level sociodemographic variablesdeducation, age and population densitydmeasured at the NUTS-3 level that others have found important in determining political participation in various contexts (Blais, 2000; Blais et al., 2004; FauvelleAymar and François, 2005; Stegmaier, 2004; Tworzecki, 2003; Wolfinger and Rosenstone, 1980). It is important to note that while the underlying theory for how these factors influence turnout is individual-level, we are using these as controls, as is generally done in aggregate turnout studies (Fauvelle-Aymar and François, 2005; Kostadinova and Power, 2007; Pacek, 1994). Education is the proportion of the population over 15 years old whose education level is above the secondary level.12 Population density is the number of thousand habitants by square kilometer in the district. Age is measured as three separate variables. First is the proportion of the population aged 20 or older that falls between 20 and 29 years old. 13 Second is the proportion of the population 20 or older who are between 30 and 59 years old. And finally, we measure the proportion of the population 20 or older who are over 60 years old.14 This categorization allows us to isolate the younger and older people, who typically participate at different rates than middle-aged people. In Fig. 5 we present a diagram of the factors that influence EP turnout. The local factors, measured at the

12 We start with 20-year-olds rather than 18-year-olds for a purely technical reason. In most censuses, the population is reported in 5-year intervals. The population between ages 15 and 19 is the interval before the interval we start with, and since many people in this group are ineligible to vote, we do not include this group. 13 The sum of these three population variables is equal to one. 14 When the White’s method is not used, the value of the estimated standard errors for the individual coefficients does not vary much, which suggests that heteroskadasticity is not really an issue in our data.

667

Table 2 Summary statistics for the explanatory variables Number Mean of regions Education Pop 20–29 Pop >60 Pop density Time to next election Number of elections EU referendum dyes vote GDP per capita Unemployment rate

119 119 119 119 119 119 119 119 119

0.14 0.20 0.25 0.26 516.27 1.52 80.73

Std. Dev. Min. 0.12 0.02 0.30 0.62 261.68 1.04 9.32

5753.41 2886.26 13.14 6.91

0.05 0.14 0.19 0.01 0.00 1.00 48.43

Max. 0.76 0.25 0.38 3.33 991.00 4.00 96.09

2025.90 17365.60 3.80 30.90

NUTS-3 level include economic measures, EU support, and the socioeconomic variables. The political structure variables are the national-level factors that influence turnout. 4. Results In Table 3, we present a series of OLS regression models of EP turnout: the economic model, the political model, the political-economy model, and the comprehensive model. Since a pooled cross-sectional model may violate the homoskedasticity assumptions of the OLS model, the estimated test of statistical significance for individual coefficients (the Student t) are based on White’s heteroskedastic robust standard errors (White, 1980).15 Our expectation regarding the impact of the economy on voter turnout was that turnout would be lower in regions with worse economic conditions. If one refers to Figs. 2 and 3, the inverse relationship between unemployment and turnout can easily be detected. Regions with higher unemployment typically have lower turnout rates. This suggests that our hypothesis holds waterdat least at the bivariate level. When we look at the Table 3, Model 1, we indeed see evidence of political withdrawal in regions with poor economic conditions, though the only economic factor that seems to really matter for turnout is unemployment.16 Looking at the coefficient on the unemployment rate variable, we see that a one percentage point increase in the unemployment rate results in essentially a one percentage point decrease in the regional turnout rate. The economic model by itself explains 42% of the variance in turnout rates. But, of course there is more to the story. That unemployment emerges as the single statistically significant economic measure meshes neatly with the findings of Pacek (1994) and Kostadinova (2003). Both Pacek and Kostadinova assessed the impact of economic conditions on turnout in Central and Eastern Europe. Pacek used only the unemployment rate as an economic measure in his turnout models and found that higher district-level unemployment produced lower turnout. Kostadinova (2003) tested the

15 We tried other measures of economic growth (such as the rate of growth of the GDP per capita) and growth over a longer period (such as the rate of growth over the period 1999–2002), but none of these proved to be significant. 16 But it is highly significant in the other regressions that we run.

668

C. Fauvelle-Aymar, M. Stegmaier / Electoral Studies 27 (2008) 661–672

Fig. 4. The percentage of the ‘‘yes’’ vote in the European referendum.

withdrawal hypothesis by using national-level time-series data and found no statistically significant effect, but her analysis used GDP per capita growth and inflation rather than a measure of unemployment. Turning to Model 2 in Table 3, the results concerning the political variables are all as expected. Together, the political variables account for less variance in turnout than the economic variables, but their effects are not negligible. The political variables explain 16% of the variance. The ‘‘Yes vote’’ variable is the percent of voters who voted ‘‘yes’’ in the referendum to join the EU. We use this as a measure of support for the EU and hypothesized in H2 that turnout would be higher in regions with higher EU support. We see that this relationship holds as the coefficient associated to the ‘‘Yes vote’’ variable is positive and statistically significant. A one percentage point increase in the referendum ‘‘yes’’ vote produces a one-third of a percentage point increase in the regional turnout rate. Our finding that the level of EU support matters for turnout sets our results apart from some other analyses of EU turnout. It is an important result in the context of the new member states because many observers interpreted the low EP election turnout as a sign that citizens in the postcommunist states, who had overwhelmingly decided to join the EU in 2003, had suddenly rejected the EU. Our finding shows that this is not necessarily the case. Regions with strong support for the EU in the referendum also produced higher turnout in the EP election. Conversely, in regions with a smaller ‘‘yes’’ vote, turnout was lower. Why have we found that EU support matters, while some others have failed to find a relationship? One possibility, suggested above, is that support for the EU may matter more in the post-communist states than in the established member states, because of the communist

political and economic legacies. For some citizens, membership in a European economic and political institution is the antithesis of their lifelong beliefs. While some citizens in the established member states may hold the same views, a larger share of the post-communist populations may hold these views, and hold them strongly. This may then be a primary cause of abstention. A second possibility is that the use of regional-level data instead of nationallevel data to test for the influence of EU support on turnout is more appropriate, since a national-level EU support indicator may conceal high variance inside countries. The coefficient associated with the ‘‘Time to next election’’ variable is negative and just attains statistical significance.17 Since this variable is measured in days, the coefficient is tiny, but one can interpret it as a one day increase until the next general election produces a 0.006 percentage point drop in EP turnout. Thus, as we expected, in countries where the next general election is held soon after the EP election, turnout is higher. Conversely, in countries where the next general election will be held in the distant future, turnout is lower. This indicates that Central and East European voters use the EP elections as a mechanism for sending information to political actors as hypothesized earlier. Since the Lithuanian case presents us with a concomitant election, it is captured in the ‘‘Time to next election’’ variable as a same-day election. In separate tests not presented here, a Lithuania country dummy variable is added

17 When the ‘‘time to next election’’ variable is excluded, the Lithuania dummy variable becomes highly significant. This result is not shown in the table to save space, but it can be obtained from the authors.

C. Fauvelle-Aymar, M. Stegmaier / Electoral Studies 27 (2008) 661–672

National Factors

Turnout in the

Number of elections

European

Time to next general

Parliamentary

elections

669

Elections

Local Factors

Socio-economic variables

Economic variables

Political variable

Age of the population

Unemployment rate

EU support

Level of education

Economic growth rate

Population density

GDP per capita

Fig. 5. Diagram of factors influencing EP turnout.

to the model but is not significant.18 This means that there is no reason other than the concomitance of elections that explains the high level of turnout in Lithuania. Ceteris paribus, if the presidential election had not been held this day (the same day as the EP elections) but later, turnout would have been much lower. We can also add that the coefficient associated with the ‘‘Time to next election’’ variable remains highly significant (and negative) when the regression is run with the Lithuania observations excluded. That means that the Lithuanian concurrent election is not an extreme outlying case that forces our explanatory variable to be significant. For the last political variable in our model, ‘‘Number of elections’’, the result again is as expected (see hypothesis 5). The effect of voter fatigue is represented by the negative and highly significant coefficient in Model 2. An increase of one additional election held in the 18 months prior to the EP election results in just over a 2.5 percentage point decrease in the turnout rate. Model 3 in Table 3 introduces a combined politicaleconomy model of turnout in the EP elections. Since only the unemployment variable attained statistical significance in Model 1, we exclude the other economic measures in Model 3. As we would expect, this combined model explains well over half the variance in turnout, with an adjusted-R2 of 0.63. All variables retain their statistical significance and coefficient sign. Further, while the values of the coefficients change slightly, the substantive impacts on turnout remain generally the same.

18 Since the population variables are each a percentage of the population age 20 or over that falls into the age category, we cannot include all 3 variables at once. If we did, we would have perfect multicollinearity (Lewis-Beck, 1980). Also, we have run many different regressions and the results repeatedly show that regions with greater proportions of older people, have higher turnout rates.

In Model 4, we present a comprehensive model of turnout that controls for some standard demographic factors that others have found to be important in explaining turnout rates. In this model, over 81% of variance is explained. Further, our hypotheses about the role of the economy and political factors in shaping turnout endure this rigorous testdthis is true when it comes to the direction of the relationship and the level of confidence. The coefficient on the unemployment variables shows us that a one percentage point increase in the regional unemployment rate produces a decrease in regional turnout by just over sixtenths of a percentage point. The ‘‘Yes vote’’ and ‘‘Time to next election’’ variables have about the same impact as in Model 2, but the ‘‘Number of elections’’ variable has a stronger substantive impact. In Model 4, the increase of one additional election in the 18 months prior to the EP election results in a decrease in turnout by nearly 5 percentage points. This model confirms that economic conditions, specifically unemployment, and political circumstances matter for turnout in the EP elections. The coefficients on the demographic variables show both expected and surprising results. First, population density fails to attain statistical significance, and thus we can only say that this appears to have no effect on regional turnout rates. Second, education performs as normal. In regions where larger proportions of the population have high levels of education, turnout is higher. The coefficient on the variable is 17.435, which seems large; however, the variable is measured as a proportion of the population where the mean is 0.14 (see Table 2). For ease of interpretation, we can move the decimal points two places to the right and think of this as a percentage. Thus, the interpretation of the coefficient is a one percentage point increase in the population with post-secondary education leads to a 0.17 percentage point increase in turnout. Third, the results concerning the age structure of the population are

670

C. Fauvelle-Aymar, M. Stegmaier / Electoral Studies 27 (2008) 661–672

Table 3 Voter turnout in the 2004 European Parliamentary elections Independent variables

Model 1

Unemployment rate GDP per capita GDP growth

0.96*** (7.57) 0.001 (1.52) 6.13 (0.83)

Yes votedEU referendum Time to next election Number of elections

Model 2

Model 3

Model 4

1.02*** (14.48)

0.319*** (2.67) 0.006* (1.69) 2.63*** (3.67)

0.229** (2.24) 0.009*** (3.26) 4.05*** (6.51)

Education Pop 20–29 Pop >60 Pop density Constant Adjusted R N

42.83*** (9.77) 2

0.42 119

9.47 (0.96)

33.87*** (3.93)

0.16 119

Beta coefficients

Model 5

0.64*** (8.46)

0.441

0.208*** (3.85)

0.401*** (5.01) 0.007*** (3.80) 4.73*** (13.31)

0.374 0.191 0.493

17.436*** 20.29 139.05*** 0.19

0.209 0.04 0.415 0.012

(4.04) (0.39) (3.65) (0.23)

26.19 (1.16)

0.63 119

0.81 119

0.059* (1.73)

23.88*** 76.96 54.72*** 2.14***

(4.22) (0.86) (3.05) (4.69)

Country dummies 0.96 119

Student t-statistics (corrected through the White (1980) method) are given in parentheses. ***p < 0.01; **p < 0.05; *p < 0.10. Dependent variable: level of turnout in the NUTS-3 regions of the 8 new EU post-communist countries.

intriguing, because they do not fit the typical trend with turnout. Our results show that turnout is higher in regions with the largest proportion of older people. In Model 4, the coefficient on the ‘‘Pop >60’’ variable is 139.05. Just like with the education measure, the population variables are measured as a proportion. The average proportion of residents over 60 is 0.25 (see Table 2), which can be translated as 25% of the population aged 20 and over are people above the age of 60. Thus, for ease of interpretation, we can say that a one percentage point increase in the population over aged 60 produces a 1.39 percentage point increase in turnout. In the column next to Model 4, we present the standardized regression coefficients to assess which factors matter most in determining turnout. The impact of election fatigue, represented by the number of elections, just edges out unemployment as the most important determinant of turnout, with standardized coefficients of 0.493 and 0.441 respectively. Recent opportunities to express views in national-level elections deflate voters’ energy to return to the polls yet again. But yet, while this political variable matters greatly in explaining turnout, the economic context matters as well. The impact of unemployment exceeds the effects of ‘‘Time to the next election’’, the referendum ‘‘yes’’ vote, and the individual socio-demographic variables as indicated by standardized coefficients. To finish, we run some robustness tests. First, we included country dummy variables in order to take into account factors that may influence turnout differently in each country. The introduction of these dummy variables necessitates the exclusion of two explanatory variables from our regression, ‘‘Time to next election’’ and ‘‘Number of elections’’, because these variables are defined at the national level. If we were to leave them in the model, there would be perfect multicollinearity between these two variables and the country dummy variables. The result of the regression is presented in Table 3 (Model 5). Except for the population density variable whose coefficient is now positive and highly significant, the other relationships continue to hold. However, the introduction of these country

dummy variables logically increases the value of the adjusted-R2 to 0.95 compared to 0.81 in Model 4. To ensure that our results were not driven by some outliers or influential observations, we calculated the studentized residuals and excluded the observations with an absolute value of their studentized residual superior to 1.96. Our results still hold with no significant changes in either the value of the coefficients or their statistical significance. In sum, we can be fairly confident about the validity and robustness of our econometric results. 5. Conclusion Our central theoretical argument in this paper, that the economic and political contexts affect voter turnout, is not new; however, our adaptation of the theory to the postcommunist environment and the application of these hypotheses in the context of EP elections at the regional level in these new EU member states is. The impact of the economic transformation over the last 15 years in Central and Eastern Europe has accentuated economic differences within countries. The change has been rapid and difficult for certain areas and their residents. As expected from our hypothesis (and confirming the work of Rosenstone (1982)), in the regions with depressed economies, especially measured through high unemployment rates, participation in the EP elections was low. Conversely, in regions with many job opportunities, voters turned out at higher rates. The overall importance of unemployment in explaining turnout speaks to the need for scholars to include this factor in studies of turnout and voter participation in national and EP election studies in the post-communist countries. With regard to our political hypotheses, national-level measures of frequency of elections and time until the next election proved important in explaining turnout rates as we expected based on previous research. Further, our analysis confirmed the theoretical argument posed by us and others, that in regions with higher levels of EU support, turnout should be higher. Still, our result is distinctive, as

C. Fauvelle-Aymar, M. Stegmaier / Electoral Studies 27 (2008) 661–672

others have failed to find evidence that EU support matters for EP turnout at the aggregate level. A part of this finding stems from our distinctive measure of support. Previous aggregate-level studies that included a measure of EU support used aggregated survey responses (Flickinger and Studlar, 2007; Franklin et al., 1996; Mattila, 2003), but consistent survey data are not available currently at the NUTS-3 level across these countries. However, because the EU referenda were held in the year prior to the EP elections, and the referenda results were available at the NUTS3 level, we were able to use this measure to assess the relationship between EU support and turnout at the regional level. This is a one-time opportunity. By the time the next EP election occurs, the referenda results will be too outdated to be a reliable measure of an area’s level of EU support. What do our findings mean for the future study of EP turnout? While our study provides an initial benchmark for understanding the impact of economic and political conditions on turnout in EP elections in post-communist countries, there are reasons to anticipate that work on later elections might uncover dynamic relationships. First, there is the question of whether or not the impact of the economy and EU support are transitory in the post-communist countries. It may be that the significant and disruptive economic transformations have amplified the withdrawal effect and that in the future, the regional economy might not have such a strong impact on turnout relative to other factors. Second, the political debates and referenda on ‘‘returning to Europe’’ may have heightened awareness and strengthened feelings about the EU to the point that it affected turnout in these countries’ first EP election. Over time, awareness of and sentiments about politics in Brussels may fade. This may lead to lower turnout, or produce changes in the primary determinants of turnout. Finally, as the costs and benefits of EU membership and EP decisions are felt by the citizens and regions, this might alter turnout in ways that correspond to their experiences. Continued study of future EP elections, at the aggregateand individual-levels, will enable scholars to more fully understand the impact of EU integration and economic transformation on electoral participation. Acknowledgments We thank Bernard Lortic for assistance in creating the maps. Previous versions of this paper were presented at the 2006 Southern Political Science Association, 2006 Midwest Political Science Association Meeting, 2006 Virginia Social Science Association, 2007 "Life in Motion: Shifting Spaces, Transcending Times, Crossing Borders" conference at Masaryk University, Brno, Czech Republic. We are grateful for the thoughtful feedback from these conference participants and the helpful suggestions from the anonymous Electoral Studies reviewers. References Aarts, K., Wessels, B., 2005. Electoral turnout. In: Thomassen, J. (Ed.), The European Voter: A Comparative Study of Modern Democracies. Oxford University Press, New York, pp. 64–83.

671

Adshead, M., Hill, J., 2005. Elections to the European Parliament, June 2004: The 15 established member states. Electoral Studies 24, 537–545. Auers, D., 2005. European elections in eight new EU member states. Electoral Studies 24, 747–754. Bell, J., 1997. Unemployment matters: voting patterns during the economic transition in Poland, 1990–1995. Europe-Asia Studies 49 (7), 1263–1291. Blais, A., 2000. To Vote or Not to Vote: The Merits and Limits of Rational Choice Theory. University of Pittsburgh Press, Pittsburgh. Blais, A., Gidengil, E., Nevitte, N., Nadeau, R., 2004. Where does turnout decline come from? European Journal of Political Research 43, 221–236. Blazyca, G., 2003. Managing transition economics. In: White, S., Batt, J., Lewis, P.G. (Eds.), Developments in Central and East European Politicsd3. Duke University Press, Durham, NC, pp. 213–233. Blondel, J., Sinnott, R., Svensson, P., 1997. Representation and voter participation. European Journal of Political Research 32, 243–272. Christin, T., 2005. Economic and political basis of attitudes towards the EU in Central and East European countries in the 1990s. European Union Politics 6 (1), 29–57. Cox, T., 2003. Changing societies: class and inequality in Central and Eastern Europe. In: White, S., Batt, J., Lewis, P.G. (Eds.), Developments in Central and East European Politicsd3. Duke University Press, Durham, NC, pp. 234–252. Delwit, P., 2002. Electoral participation and the European poll: a limited legitimacy. In: Perrineau, P., Grunberg, G., Ysmal, C. (Eds.), Europe at the Polls: The European Elections of 1999. Palgrave, New York, pp. 207–222. van der Eijk, C., Oppenhuis, E., 1990. Turnout and second-order effects in the European elections of June 1989d Evidence from the Netherlands. Acta Politica 25 (1), 67–94. Fauvelle-Aymar, C., François, A., 2005. Campaigns, political preferences and turnout: an empirical study of the 1997 French Legislative Elections. French Politics 3 (1), 49–72. Fauvelle-Aymar, C., Lafay, J.-D., Servais, M., 2000. The impact of turnout on electoral choices: an econometric analysis of the French case. Electoral Studies 19, 393–412. Fidrmuc, J., 2000. Economics of voting in post-communist countries. Electoral Studies 19, 199–217. Franklin, M.N., 2001. How structural factors cause turnout variations at European Parliament elections. European Union Politics 2, 309–328. Franklin, M.N., 2002. The dynamics of electoral participation. In: LeDuc, L., Niemi, R.G., Norris, P. (Eds.), Comparing Democracies 2: New Challenges in the Study of Elections and Voting. Sage Publications, Thousand Oaks, CA, pp. 148–168. Franklin, M.N., 2004. Voter Turnout and the Dynamics of Electoral Competition in Established Democracies Since 1945. Cambridge University Press, Cambridge. Franklin, M., van der Eijk, C., Oppenhuis, E., 1996. The institutional context: turnout. In: van der Eijk, C., Franklin, M.N. (Eds.), Choosing Europe? The European Electorate and National Politics in the Face of Union. University of Michigan Press, Ann Arbor, MI, pp. 306–331. Flickinger, R.S., Studlar, D.T., 2007. One Europe, many electorates? Models of turnout in European Parliament elections after 2004. Comparative Political Studies 40, 383–404. Geys, B., 2006. Explaining voter turnout: a review of the aggregate-level research. Electoral Studies 25, 637–663. Glass, D., Squire, P., Wolfinger, R., 1984. Voter turnout: an international comparison. Public Opinion 6 (6), 49–55. Hix, S., Marsh, M., 2007. Punishment or protest? Understanding European parliament elections. Journal of Politics 69, 495–510. Jackman, R.W., 1987. Political institutions and voter turnout in the industrial democracies. American Political Science Review 81, 405–423. Jackman, R.W., Miller, R.A., 1995. Voter turnout in the industrial democracies during the 1980s. Comparative Political Studies 27, 467–492. Jurkynas, M., 2005. The 2004 presidential and parliamentary elections in Lithuania. Electoral Studies 24, 741–784. Kostadinova, T., 2003. Voter turnout dynamics in post-communist Europe. European Journal of Political Research 42, 741–759. Kostadinova, T., Power, T.J., 2007. Does democratization depress participation? Voter turnout in the Latin American and Eastern European transitional democracies. Political Research Quarterly 60, 363–377. Koepke, J.R., Ringe, N., 2006. The second-order election model in an enlarged Europe. European Union Politics 7 (3), 321–346. Landesmann, M.A., Rosati, D.K., 2004. Shaping the New Europe: Economic Policy Challenges of European Union Enlargement. Palgrave Macmillan, New York. Lavigne, M., 1999. The Economics of Transition: From Socialist Economy to Market Economy, second ed. St. Martin’s Press, New York.

672

C. Fauvelle-Aymar, M. Stegmaier / Electoral Studies 27 (2008) 661–672

Lewis-Beck, M.S., 1980. Applied Regression: An Introduction. Sage University Press, Newbury Park, CA. Lewis-Beck, M.S., Lockerbie, B., 1989. Economics, votes, protests. Comparative Political Studies 22 (2), 155–177. Lewis-Beck, M.S., Stegmaier, M., in press. The economic vote in transitional democracies. Journal of Elections, Public Opinion and Parties. Marsh, M., 1998. Testing the second-order election model after four European elections. British Journal of Political Science 28, 591–607. Mattila, M., 2003. Why bother? Determinants of turnout in the European elections. Electoral Studies 22, 449–468. McCormick, J., 2005. Understanding the European Union, third ed. Palgrave Macmillan, New York. Nugent, N. (Ed.), 2004. European Union Enlargement, fifth ed. Palgrave Macmillan, New York. Pacek, A., 1994. Macroeconomic conditions and electoral politics in East Central Europe. American Journal of Political Science 38, 723–744. Radcliff, B., 1992. The welfare state, turnout, and the economy: a comparative analysis. American Political Science Review 86, 444–454. Reif, K., Schmitt, H., 1980. Nine second-order national elections: a conceptual framework for the analysis of European elections results. European Journal of Political Research 8, 3–44. Roland, G., 2000. Transition and Economics: Politics. Markets and Firms. MIT Press, Cambridge, MA. Rose, R., 2004. Europe expands, turnout falls: the significance of the 2004 European Parliament election. www.idea.int/elections/voter_turnout_europe/EU_Elections_june04.pdf.

Rosenstone, S.J., 1982. Economic adversity and voter turnout. American Journal of Political Science 26, 25–46. Schmitt, H., 2005. The European Parliament elections of June 2004: still second-order? West European Politics 28 (3), 650–679. Schmitt, H., Mannheimer, R., 1991. About voting and non-voting in the European elections of June 1989. European Journal of Political Research 19, 31–54. Stegmaier, M.A., 2004. Economic evaluations and electoral participation in Poland, Hungary, Slovakia, and the Czech Republic. Virginia Social Science Journal 39, 71–90. Tucker, J.A., 2006. Regional Economic Voting: Russia, Poland, Hungary, Slovakia and the Czech Republic. Cambridge University Press, New York. 1990–1999. Tucker, J.A., Pacek, A.C., Berinsky, A.J., 2002. Transitional winners and losers: attitudes toward EU membership in post-communist countries. American Journal of Political Science 46, 557–571. Tverdova, Y.V., Anderson, C.J., 2004. Choosing the West? Referendum choices on EU membership in east-central Europe. Electoral Studies 23, 185–208. Tworzecki, H., 2003. Learning to Choose: Electoral Politics in East-Central Europe. Stanford University Press, Stanford, CA. Vachudova, M.A., 2005. Europe Undivided. Oxford University Press, Oxford. Vecˇernı´k, J., Mateˇju˚, P. (Eds.), 1999. Ten Years of Rebuilding Capitalism: Czech Society after 1989. Academia Press, Prague. White, H., 1980. A Heteroscedasticity-consistent covariance matrix estimator and a direct test for heteroscedasticity. Econometrica 48, 817–838. Wolfinger, R.E., Rosenstone, S.J., 1980. Who Votes? Yale University Press, New Haven, CT.