Local Effects in Local Elections General Conference, Prague

Transcription

1 Local Effects in Local Elections 2016 General Conference, Prague Panel: Local Democracy and Local Self-Government: Coping with Challenges in Times of Change Michal Pink, Ph.D. and Petr Voda, Ph.D. Department of Political Science International Institue of Political Sciences of Masaryk University Faculty of Social Studies Masaryk University Brno, Czech Republic THIS PAPER IS A WORK IN PROGRESS. Abstract In this paper we analyse local elections in the city of Brno, Czech Republic, by using multilevel regression models and discuss the explanatory potential of indicators on both individual and aggregate levels when it comes to electoral behaviour. Our research focuses on how environment affects political behav9iour of individual voters. More precisely, the analyses shows how the individual level factors influencing the intention to vote and party selection in local elections interact with factors widely used in electoral geography as indicators of Rokkan s and Lipset s cleavage theory (e.g. unemployment, number of entrepreneurs, relation centre-periphery etc.). The set of individual level factors covers political interest, knowledge and efficacy, social class and sense of responsibility. This perspective allows us to eliminate some of the weaknesses of conclusions drawn from aggregate data. The case of Brno is selected because of the availability of data suitable for the purpose of the analysis. Using any national sample, the additional spatial information can be attributed only to very large and internally heterogeneous regions. In this example, the data are based on local surveys in Brno; an exit poll conducted during the local elections in 2014, and a survey conducted in 2015 on a random sample of all inhabitants of the city. The spatial information is therefore linked to relatively small and compact neighbourhoods. Introduction In recent times, analyses in the field of voting behavior and electoral geography have brought numerous important and interesting information into political science. First of all, we have several theoretically distinctive explanations of voters decisions and also explantions why are electoral results different in different regions of countries. However, these explanations are almost always isolated from the perspective of the other level. Scholars analyzing voting behavior usually do not count with the fact that places produce incentives for certain electoral

2 behavior (see e.g. Johnston and Pattie 2004), and similarly scholars in electoral behavior often omit the individual base of aggregated electoral results. The idea that findings of analysis made on different conceptual levels can be connected in multilevel models is not new. Jones at al. came with this approach already in Using data for the 1987 British general election, they analyzed how individual s decisions are influenced by constituency and region level variables. Very similar procedure was followed also by Steenbergen and Jones (2002) in their article introducing multilevel models into political science. However, since the presentation of their idea, only few studies actually using this approach has been conducted. Moreover, most of such studies link the contextual effects to level of states and these studies are very popular especially when electoral turnout or trust in institutions is analyzed, because it is easy to have equivalent meaning of values of dependent variable in different countries. Shane Singh (2011) analyzed whether individuals living under compulsory voting rules are more likely to go to the polls when penalties for abstaining are both strict and routinely enforced. He used a sample of 36 countries, some with compulsory voting and some with voluntary rules. Using surveys conducted in 20 European democracies, Anderson and Singer (2008) examine the effect of income inequality on people s attitudes about the functioning of the political system and trust in public institutions. Fieldhouse et al. (2007) analyses variations in turnout of young people in 22 European countries and how it is influenced by their characteristics and to what extent by the electoral context in each country using 2002 ESS data. Both turnout and vote choice is analyzed in the article by Ghitza and Gelman (2013), however, geographical location is the only item in their grouping variable. Duch and Stevenson (2008) test a hypothesis whether sharing policy-making responsibility more equally among parties leads to higher salience of economic considerations in vote decision. They use data from 163 national surveys from 18 countries over a 22-year period. Only a few texts analyze voting behavior in terms of selection or preference of a certain party. Petr Voda (2015) examines the effect of post-communism (and several aspects of political system like proportionality or polarization and fragmentation of the party system) on value and class voting in 20 European countries. The focus of this paper is also related to subject of studies from other fields which work with individual behavior influenced by small geographical units. Kalff et al. (2001) examined whether neighbourhood level socioeconomic variables have an independent effect on reported child behaviour problems over and above the effect of individual level measures of socioeconomic status. They use multilevel analysis of cross sectional survey data relating individual level child behavioural problems and parental measures of socioeconomic status to neighbourhood level measures of socioeconomic deprivation in the city of Maastricht, the Netherlands. Islam et al. (2006) test the hypothesis whether individual health status is related to area-level social capital measured by rates of voting participation in municipal political elections, controlling for personal characteristics. Their analysis is based on panel data from Statistic Sweden's Survey of Living Conditions (the ULF survey). Dataset includes 24,419 individuals nested within 275 Swedish municipalities.

3 In our study, we do not focus on big (parliamentary) elections in large units (countries). Instead, we focus on local elections in one city (Brno) in local units (neighbourhoods). Local politics and elections in the Czech Republic tend to be presented as based on personalities rather than ideology (Šedo 2009). Eg., one of Brno s previous city mayors, Roman Onderka, was knowen for claiming that there are no leftist or rightist solutions in local politics, only bad and right solutions based on a wide consensus and apolitical cooperation between citizens. Our genaral research question is about the extent in which are voters choices in local elections actually based on the ideological preferences. Furthermore, we also analyzed what kind of local environment weakens or stregtens effects of ideology on vote choice. Local Elections in Brno Before we move to theory and analysis, i tis neccesary to introduce the basic outline of electoral system, because it makes in the case of czech local elections very specific conditions for voting behavior. Generally the system is based on the proportional election system with candidate lists. The d Hondt method is used to allocate mandates between candidate lists, with a 5% administrative threshold for candidate lists with full number of candidates. The threshold is lower for lists with fewer candidates. The difference from other proportional systems used in the Czech Republic (and most other countries) is that each voter has a number of votes equal to the number of elected representatives in the municipality. Voters have three options how to use their votes. Firstly, the voters can choose one candidate list without selecting individual candidates which means giving all votes to this one list. Secondly, they can also choose individual candidates across the candidate lists and thirdly, they can combine these two options. When selecting a candidate, the voter simultaneously gives one vote to the whole list (something a lot of voters do not realize) and give a preferential vote for the candidate. With sufficient number of preferential votes, a candidate can move up on the list. The necessary quorum for vertical shifts within the candidate list is 110 percent of the preferential votes average within the given candidate list. What is important her eis the fact, that voter can vote for different parties, which has some important consequences, especially for our analysis as is discussed bellow Elections in Brno Brno is the second largest city in the Czech Republic and one of four large metropolitan areas, with a population of about half million, including the suburbs. Until 2006, the local elections had been won by the liberal conservative party ODS 1, which also controlled the office of the mayor. In 2006, ODS again won the elections in Brno, but the subsequent negotiations resulted in a center-left coalition. The city was ruled by a social democratic mayor for the next eight year until the 2014 elections, with the ODS as minor coalition partner from For English and Czech names of the parties see table 1.

4 There were over 30 candidate lists competing in the 2014 elections and eight that reached the 5% threshold. Among them were the two major parties and incumbent coalition members, the ČSSD (holding the post of the city mayor), and the ODS. The existing opposition parties were joined by two new formations: the nation-wide populist party ANO 2011 (a governmental party at the national level in 2014) and the local protest party Žít Brno formed around activist Matěj Hollan. The winner of the 2014 elections became ANO 2011 which gained just under 20%. ČSSD came in second and Žít Brno third. The full list of successful parties and their electoral gains can be seen in table 1. Table 1. Electoral results local elections Brno 2014 Party candidate list % of votes Number of seats KSČM 6,73 4 The Communist Party of Bohemia and Moravia Komunistická strana Čech a Moravy ČSSD 17,67 11 The Czech Social Democratic Party Česká strana sociálně demokratická KDU-ČSL 11,83 7 The Christian and Democratic Union Czechoslovak People's Party Křesťanská a demokratická unie Československá strana lidová Žít Brno 11,89 7 Live Brno Žít Brno SZ 7,38 4 The Green Party Strana zelených ODS 7,62 5 The Civic Democratic Party Občanská demokratická strana TOP 09 6,57 4 ANO ,92 13 Total 55 Source: Data We are using data from three sources - a) 2014 Brno local elections exit-poll, b) 2011 census data and c) electoral districts in map layers. An exit-poll is a poll of voters taken as they leave the polling stations, therefore during the actual elections. In our case, the population were all voters who took part in the municipal elections (to the General Assembly of the City of Brno, not the assemblies of city districts). To obtain a representative sample, voting stations were chosen systematically and voters were

5 chosen randomly within the selected voting stations. The poll ran throughout the elections, from the time the voting stations opened on October 10 th till the time they closed on October 11 th. We have obtained answers from 688 respondents from 41 voting stations. Unfortunately the voting districts (areas covered by one voting station) are not the same as any administrative units dividing the city, therefore the locations had to be aggregated for the data to be used in multilevel analyses. Therefore, the final data includes 688 respondents belonging to 34 different territorial units. The data was weighted only in respect to the voter turnout in each territorial unit, since there are no other characteristics known about the population of voters who took part in the elections. Due to the character of the electoral system, it is now even possible to weight the data according to the actual electoral results. Unless the voter chose to check only one candidate list and no individual candidates, it is virtually impossible for a survey to find out which candidates he or she selected from each party. Table 2 displays the comparison of the exit-poll data and actual electoral results. Table 2. Comparison of exit-poll data and actual electoral results Actual Confidence Interval 95% Exit-Poll Results Electoral Results Standard Error Lower Bound Upper Bound Difference ANO ČSSD KSČM KDU- ČSL ODS SZ TOP Žít Brno The exit-poll data is relatively accurate. The difference between the actual electoral results and the results predicted by the exit-poll is in most cases relatively small. In the case of ANO 2011, Žít Brno and the Green Party, the difference is minimal. In the case of ČSSD, KSČM and KDU-ČSL, the data from exit-poll was undervalued. This could be explained by respondents unwillingness to admit to voting for these parties or/and inappropriate selection of voting stations where the exit-poll was conducted. In case of the KDU-ČSL the difference could also be attributed to the popularity of individual candidates (party was more popular among voters choosing individual candidates see below for more details), however, the data does not allow us to draw any definite conclusion about the causes of the difference. Exit-poll results for TOP09 and ODS were overvalued. The exit-poll and survey data were complemented by data from the 2011 census in the Czech Republic. We used data for lowest territorial units and aggregated them to the level of territorial units used in the survey and exit-poll. To aggregate the data, map layers depicting different levels territorial units.

6 Methods To find out the answer to our questions, we employ several methods in several steps. Firstly, we use factor analysis to create our contextual variables. The analysis is run on the 2011 census data related to electoral districts. Because delimitation of electoral districts do not match the stastical units, we aggregate the districts and units to areas where the boundaris of units for which we have data about population match the boundaries of electoral districts.the outcome of the analysis are three latent variables which are labeled as owner worker (because of high positive loading of employees and people living in prefab houses and negative loading of entrepreneours and people living in family houses), industry-center (high loading of proportion of people working in industry and whit vocational education, negative loading of proportion of people with university degree) and marginalized suburb localities (negative loading of unemployment, positive loading of ownership of flat or house). Factor loading are saved and used as independent variables in following analysis. Binomial multilevel logistic regression is employed to analyze party choice. The analysis of of party choice is computed on data from the exit-poll and contextual variables described above. Since we have variables related to two conceptualy different levels individual voters and localities (aggregates of electoral districts) we estimate the coefficients in multilevel setting. Separate models are conducted for each party. The dependent variable of each analysis is a vote for the given party: 1 means vote for the party, 0 means vote for any other party. This approach poses a problem with the reference category which can have a different meaning for each party. However, since voters can vote for more than just one party, the categories are not exlusive (for example: there are voters who vote for both ANO and CSSD, voters who vote for ODS and TOP09 and so on). Left and right in local politics in Brno As we already mention in introduction, there are some doubts about the role of ideological factors in local politics, which can be reflected also in voting behavior on local level. If would be true, that politics on local level has different base than ideology then we should find a) no differences between parties regarding their position on the left-right axis or b) no differences between voters of different in their position on left-right scale. Firstly, we present the simple crosstabulations to illustrate the pattern of positioning parties and voters on left-right scale. Table 3. Voters perception of candidate lists ANO ČSSD KSČM KDU-ČSL ODS SZ TOP09 Žít Brno Left - 3,74 34, ,67 28, ,59 8,41 15, ,78 3 3,17 31,78 6,25 7,81-8,89-4,17 4 3,17 19,63-10,94 3,45 15,56-6, ,60 24,30 6,25 20,31 3,45 33,33 1,89 29,17

7 6 19,84 1,87 3,13 25,00 8,62 15,56 1,89 18, ,29 5,61 3,13 18,75 8,62 20,00 26,42 23, ,05-3,13 14,06 34,48 2,22 39,62 11,11 9 7, ,56 25,86 4,44 22,64 4,17 Right 7, ,56 15,52-7,55 - Table 3 shows where voters place individual parties on the left right scale. It is clear that even on individual level, voters discriminate between the left and the right. The distinction is stronger when it comes to older parties, with the KSČM clearly on the left and ODS and TOP09 on the right of the political spectrum. The preceptions are less clear when it comes to ANO 2011 and Žít Brno, the two relatively new protest parties. However, we can still see that they are not seen as leftist but rather as centre-right. Table 4 displays how voteters of individual parties place themselves on the same scale. The results are not suprising when compared to table 3: voters of the KSČM and ČSSD see themselves as leftist voters, voter of the TOP09 and ODS plase themselve to the right. Table 4. Voters perception of themselves ANO ČSSD KSČM KDU- ODS SZ TOP09 Žít Brno ČSL Left - 4,67 34, ,79 0,93 15, ,33-2,78 2 5,56 11,21 18,75-1,72 2,33-6,94 3 3,17 17,76 6,25 4,92 1,72 9,30-4, ,43 22,43 6,25 9,84 3,45 6,98-16, ,49 26,17 6,25 18,03-20,93 1,89 19, ,87 8,41 3,13 18,03 5,17 16,28 1,89 18, ,84 2,80 3,13 8,20 10,34 20,93 26,42 22, ,32 0,93 3,13 26,23 29,31 13,95 39,62 5,56 9 9,52 4,67 3,13 4,92 20,69 4,65 22,64 4,17 Right 0, ,20 27,59 2,33 7,55 - Results of analysis on party choice Previous bivariate analysis shows us that politics have some ideological base also in local politics, at least when comes to a big city. To make the conclusions more precise, we conduct also multivariete regression to find out the actual effect of position on left right scale on voting for different parties. Moreover, the analysis enables us to find out in which conditions is voting behavior more likely to be ideological. This chapter sumarizes the results of set of binomial multivel regressions. We present the results for each party separately, starting with leftist parties. The higher chance to vote KSČM have especially leftist voters (see Table 5). These voters have 20 times higher chance to vote KSČM than voters in centre. Moreover, it is also clear that voters who consider themselves as rightist and live in areas with high proportion of entrpereneurs find the KSČM as an

8 unacceptable choice significantly more often than rightist voters in other areas.the chance to vote KSČM is rised also by time of decision and by disatisfaction with the course where Brno is going. Voters of KSČM are mostly identified with party because those who decide at the last minute have lower chance to vote KSČM than other parties. On the contrary, with higher income the chance to vote KSČM is decreasing. An unexpected finding is that communist voters can be found more in areas with higher education. In the case of other leftist party ČSSD the chance to vote this party is incresed by the satisfaction with the situation in Brno, which is expectable result since the ČSSD was the major governing party in Brno. Not suprisingly, leftist and centrist voters have higher chance to vote ČSSD than other parties. The chance to vote ČSSD raises also with higher age and with secondary education (as opposed to primary and university education). Finnaly, we find that voters who felt that they were influenced by the electoral campaigne rather than by their opinion on the individual candidates have higher chance to vote CSSD than other parties. Leftist voters in workes neighbourhoods are 3.5 times more likely to vote for ČSSD than leftist voters in owners areas, where even leftist voters tend to vote for the ODS. Now we move to interpretation of results for right wing parties. Firstly we discuss the case of ODS. Similarly to ČSSD also in the case of ODS as former governing party in Brno, the chance to vote this party is higher among voters satisfied with situation in Brno than among dissatisfied voters (though not as much as voters of the ČSSD). In other respects are voters of the ODS very different than votars of ČSSD. The chance to vote ODS is increased by voters position on right part of left-right scale. Also voters who felt more influenced by their opinion of the candidates and by their identification with the party than by programme have higher chance to vote ODS than other parties. Business owners and students choose this party with higher chance than employees, retirees and others. The same applies for voters with middle income compared to those with low income. We also find some direct effect of contextual variables. People who live in older parts of the city and in areas with higher percentage of business owners and services have higher chance to vote ODS than other parties. Moreover, the context modifies the effect of position on left-right scale. Right-wing voters from high income areas are 5 times more likely to vote for the ODS than right-wing voters in workers areas. The second right wing party is TOP09. Similarly to ODS, right wing voters have higher chance to vote TOP09 than other parties.other variables have no or only weak effect. These variables include interest in politics, age and industrial-centre contextual variable. The chance to vote TOP09 is rising with higher interst in politics and with more central of location where voter lives and with middle category of age compared to young category. While the the right wing votershave higher chance to vote TOP09 in higher status areas than in workers areas, this factor is not as strong as for the ODS. Results for ANO 2011, KDU-ČSL and Žít Brno are not so strong as significant as results for previous parties, because this gorup of parties are rather in the centre of left-right scale and therefore the position on left-right scale do not create any difference among them.

9 The KDU-ČSL voters see themselves mostly in the right centre of the political spectrum. It is reflected also in results of regression analysis, when leftist voters have lower chance to vote KDU-ČSL than other parties than centrist and right wing voters do. The chance to vote KDU- ČSL is then increased by lower income and by emphasis on importance of social issues. Unlike with the voters of parties described above, we can say that contextual characteristics do not have significant impact on the choice to vote for the party. Other centrist parties are new in local elections. Therefore, i tis not surprising that chance vote both for ANO and ZB is higher for voter who decided few days before elections than for voters decided long time ago. The chance to vote for ANO is then increased by interest in other topics than those connected with traffic and social issues and by midlle category of age. External characteristics such as the character of the area where they live are not important in this case. Our findings are especially interested when it comes to the local protest party Žít Brno. Its voters came from all over the political spectrum. As a protest party, voters who are not satisfied with situation in Brno have higher chance to vote for the ZB than for other parties. That is probably the reason why this groups of voters cited party program proximity as their main motivation to vote for the party. The last interesting finding was that people surrounded by highly educated persons are more likely to vote for the party. Table 5: results of binomial multilevel logistic regressions for party choice leftist parties right-wing parties centre parties KSČM CSSD ODS TOP ANO KDU ZB expb sig. expb sig. expb sig. expb sig. expb sig. expb sig. expb sig. individual factors retro evaluation 0.33 * 3.50 *** 1.75 * *** political interest decision (more than 3 months before elections is reference category) 3 months - 2 days before * * less than 2 days before * left right (centre is reference category) left *** ** 0.22 ** 1.05 right *** 7.10 *** 3.11 *** issues railway station traffic * social ** 2.11 ** 0.95 transparency mobilization (personal factors are reference category) programme * campaign education (basic is reference category) high school university **

10 occupation (enterpreneur is reference category) employee * retired student other income (low is reference category) middle high * 1.57 no answer age (young is reference category) middle ** 3.08 *** 1.96 * * 0.44 ** elederly * * contextual factors " worker -owner" * "industry - center" "marginal -suburbs" * cross level interactions left * " worker -owner" right * " worker - owner" * Intercept 0.01 ** 0.05 *** 0.04 *** 0.03 *** 0.30 * 0.20 * 0.07 *** AIC variance Std.dev N=680, groups = 34 Conclusion As mentioned at the beginning, this paper is still work in progress. Tentative conclusions for our analyses are as follows: Voter choices are influenced by their environment, as well as by their socio-economic characteristics. Even in local politics, specifically in the environment of a relatively large Czech city, values (how voters place themselves on the left-right spectrum) still significantly influence voter choice. This is true for populist and protest parties such as ANO 2011 and Žít Brno as well, but to a lesser extent. When it comes to the left-right spectrum, voters preceive parties at the local level similarly as at the nation level. Therefore, local specifics do not change voters preception of parties are on the right, left, center, etc. Literature Anderson, Christopher J., and Matthew M. Singer. "The sensitive left and the impervious right: multilevel models and the politics of inequality, ideology, and legitimacy in Europe." Comparative Political Studies (2008). BALÍK, Stanislav. Local Cleavages, Politics and Policy at the Local Level Is the Depolitization Real? Sociální studia, Masarykova univerzita, 2016, roč. 13, č. 1, s

11 Duch, Raymond M., and Randy Stevenson. "Context and the economic vote: a multilevel analysis." Political Analysis 13.4 (2005): Fieldhouse, Edward, Mark Tranmer, and Andrew Russell. "Something about young people or something about elections? Electoral participation of young people in Europe: Evidence from a multilevel analysis of the European Social Survey." European Journal of Political Research 46.6 (2007): Ghitza, Yair, and Andrew Gelman. "Deep interactions with MRP: Election turnout and voting patterns among small electoral subgroups." American Journal of Political Science 57.3 (2013): Health Economics, Policy and Law (2006) 1: Cambridge University Press 2006 doi: /S X. Does it really matter where you live? A panel data multilevel analysis of Swedish municipality-level social capital on individual health-related quality of life M. KAMRUL ISLAM a1 c1, JUAN MERLO a1, ICHIRO KAWACHI a2, MARTIN LINDSTRöM a3, KRISTINA BURSTRöM a4, ULF-G. GERDTHAM a5 Jones, Kelvyn, Ronald J. Johnston, and Charles J. Pattie. "People, places and regions: exploring the use of multi-level modelling in the analysis of electoral data." British Journal of Political Science (1992): Kalff, A. C., et al. "Neighbourhood level and individual level SES effects on child problem behaviour: a multilevel analysis." Journal of epidemiology and community health 55.4 (2001): Steenbergen, Marco R., and Bradford S. Jones. "Modeling multilevel data structures." american Journal of political Science (2002): Singh, Shane. "How compelling is compulsory voting? A multilevel analysis of turnout." Political Behavior 33.1 (2011): ŠEDO, Jakub. Preferenční hlasy v komunálních volbách - zdroj obměny či stability volených reprezentantů. Acta politologica, 2009, roč. 1, č. 3. Czech Statistical Office:

12