filar 3 i iii him Survey Research I hr vole is only a very rough index of politieal attitudes, since the choices nit' highly limited. A finer scale is necessary if the nature of political atti-liulcs is to be examined more closely. Since the vote is secret, it is impos-ii ill to identify ballots and to relate directly a given kind of vote to eco- .......ii and social factors. In studying such a subject as popular interest in Voting, il is soon apparent that there are no available data on the number of i iMi' voters and on the characteristics of voters and non-voters. Ficld- ■ "i l I'v means of questionnaires and interviews is necessary to throw light ■ ■ii I'inhk-ms of this sort. H.F. Gosnell (1933: 396) inii u.inction1 i In i......mi y goal of all election surveys is to provide systematic empirical evi- i ii ili.it will help to explain election results: who won the election and why? ii ii old I. Gosnell noted in the epigraph above, official election statistics pro-pp very limited information on (a) what motivates voters to turnout to vote, and i h\ 11ic voter supported one party rather than another. Within political sci-c loot ion surveys are perhaps the most important means used to test compet- ihi-oi ics of voting behaviour (note, Evans 2004; Bartels 2010). This chapter ill imi discuss the merits of different models of vote choice in the Czech Re-i-iii'lu . as this has been discussed in detail elsewhere (see, Lebeda et al. 2007; i iii' l MHO; Lineketal. 2012). 11 "\\ over, in order to provide the reader with some sense of why election sur-iii ill their manifestations are important let us consider for a moment a con- ii ii vol ing decision: the Chamber Elections of May 28-29 2010. Rather than pre-.....In- standard theories employed by political scientists of what is likely to have ilt it i mined turnout and party choice on this occasion: we will use instead a vote flu Hi o decision-making tree constructed by concerned citizens who wished to in-;n,Nc electoral participation. This non-academic approach has the merit of dem-n.iiiiig that the explanations of electoral behaviour tend often to focus on a \ A shorter version of this chapter published in Czech is available in Krejci and Leontiyeva "i ! i luipter 10). [107] handful of key factors. According lo (he void's decision making live shown in lie ure 3.1, the main delerminants of bolh participation and parly choice in the Chain her Elections of 2010 could be summarised in terms ol loin decision nodes. First, the decision to participate was evaluated on the basis of level of inlbi inn tion and knowledge. Second, if a person understood Czech politics then the in-il decision centred on the voters' left or right wing orientation; or more concreti I supporting a party representing the 'rich' or the 'poor.' Third, once the voter hail selected their ideological orientation the main criteria for supporting a specilli left or right wing party depended on attitudes toward well-known politician, from the main left (ČSSD) or right wing (ODS) parties as appropriate. Such pol iticians were important because they were likely to hold high office in the nexl government; and facilitated making evaluations of the relative competence of lh< two coalition government alternatives. Those who are Christians had the option of either supporting the centre-rij-lil Christian Democrats (KDU-CSL); or alternatively on the basis of orientation lo ward smoking marijuana (a salient socially liberal issue) Czech voters could sc lect the more liberal Greens (SZ), or the new law and order party called Puhln Figure 3.1: The Czech voter's decision-tree in the chamber elections of 2010 Want r politici: lit, you like I iinjicr. Item. Kivc to I— And . the l>o yon smoke min-ij nana? Včel " verejne Straní zelených Do you like I'ai-uubek, Rath. S€»holka. etc? Do you like Zeman? YES-* SPO Source: http://www.motorkari.cz/forum-detail/?ft=88060&fid=34 (accessed 15/02/2012) Note that this decision-tree is based on some citizens' perceptions of the party choices that were of offm during the election campaign of May 2010. This summary description of the electoral logic confronted by all Czech voters is interesting as it demonstrates which factors were considered by mobilised citizen! most important in deciding if or how to cast a vote. ii ni (VV). fourth, il the vnlu u.r. j'l'iicially dissatisfied with the main left "I iij-lii wing parties, those on tin- richl could opt for the entry of new politi- .• es on the national stage by voting for pro-European TOP 09 (an offshoot • I I 'I ( 'SI.); or the eurosceptic economically liberal, but socially conservative 1 in/ens' Party (Strana svobodných občanů).2 Iln central lesson to be taken from the decision logic evident in Figure 3.1 is lliHl electoral behaviour is seen by politically engaged citizens to be shaped by i mlin malion and knowledge, (2) left-right ideological orientation, (3) affective in nnlalion toward party leaders or perceptions of their competence, and (4) llHH'llic issue based motivations relating to law and order, the EU, etc. The of-i 11 • In lion results of May 2010 provide little information on the relative im- , .....i c of each of these four motivations in determining the dramatic election ni. nun- observed (but see, Linek et al. 2012). Therefore, in order to understand i. . In loral democracy works in the Czech Republic election surveys are un- I. ii.il in lo measure voters' preferences and motivations. I hi mapping out of Czech election surveys in this chapter is presented as foliu 11 ic first section, a brief overview of the origins and development of mass ni ■ \nii- methods in Czechoslovakia (1967-1989) to examine political ques- II. .i i will be presented. The goal here is to show that the study of citizens' politi-i iililudes in the Czech Republic has a long history notwithstanding the retarding effects of communism where surveying was denounced as an ideologically IMixonnil form of empirical social research. Sections 2 and 3 outline the survey iii i 11 'i lower and upper chamber elections over the last two decades. This is followed by a discussion of the two European Election Studies undertaken in the 1 reh Republic since EU accession in 2004. Section 4 maps out the survey data inundated with elections to the local and regional levels of governance; and this . ii .I low lxI by an overview of exit poll results. Thereafter, the focus moves in the i M. .w ing two sections away from the immediate context of elections to the evo-i >n'ii of vote intentions and partisanship during inter-election periods as meas-.....I in panel and repeated cross-sectional surveys. In the penultimate section, III. it is an overview of the analysis of Czech electoral data from 1920 to 2010. i In r aggregate data are invaluable for studying spatial and cross-time trends in l.. loral behaviour. Thereafter, there are some concluding comments. Hi lore embarking on the mapping out of election data and associated analy-■.oine words are in order regarding immediate post-election processes such ii'i government formation and duration. This field of research while the subject i "iisiderable commentary in media and academic publications has not been Mi.ma svobodných občanů has been linked in the Czech media with the neo-liberal policies ii'imd by President Václav Klaus (founder of ODS). [108] [109] theory, D.H.i .mil Analysis examined in the same detail using the standard formal and empirical models of political science. The data for such work is available in Müller, Fellelschoss ami Harfst (2004) and Ryals Conrad and Golder (2010). For more details sec inlci net resources in the appendix. Recent research on government coalition bargain ing and government duration is given in Nikoleyni (2003), Druckman and Roh erts (2005) and Somer-Topcu and Williams (2008).3 3.1 Chamber Elections (1990-2010) The first election survey undertaken in Czechoslovakia following the Velvet Rev olution was a pre-election poll fielded between April 28 and May 11 1990.4 This survey used face-to-face interviewing with a stratified representative sample <>i the adult population. The fieldwork was undertaken by AISA and STEM and il yielded 2,710 interviews and a very high response rate of 93%.5 This extensive political attitudes and behaviour survey explored four key themes: perceptions of political parties and vote intentions, electoral participation, attitudes towards democratic elections, and attitudes toward proposals for economic and social reform (Gabal, Bogusak and Rak 1990). This pre-election survey was unique in that the poll results were designed to be used by all parties competing in the Federal Elections of June 1990 to mo bilise electoral participation. Consequently, the results of the AISA survey ol April-May 1990 were published in the national (independent) newspaper LidovM noviny in a series of articles during May and June. These data have been deposited in the German Social Data Archive (GESIS).6 One of the most comprehen sive accounts of this election, which uses this survey, is given in Klingemann (1996). A follow up post-election survey was fielded during the first half of November 1990 and it explores a wide range of domestic and international polili cal issues (Toka 2000: 152). These data are also available from GESIS (Study ZA2561). 3 Research on government formation during the First Republic is primarily historical in n.i ture. The application of standard coalition formation models from political science to Czecho Slovak, Czech and Slovak governments since 1920 is an area of research that holds considerable promise. 4 An inventory of all elections since 1990 is given in Appendix 3.1. 5 The National Democracy Institute (NDI) an influential American NGO provided AISA with expert consultation advice on the design of this survey in March 1990. For more details son, http://www.ndi.org/files/1379 sk elec.pdf (accessed 15/02/2012). 6 For details of this survey which is catalogued as 'Study ZA 2562: Czechoslovakian 1990 Parliamentary Election' and further details are available at: http://info1.gesis.Org/dbksearc:hl I sdesc2.asp?no=2562&db=e&doi 10.1232/1.2562 (accessed 15/02/2012). i lection Survey Research With the dissolution of the Czechoslovak Federal Republic in 1993, all sub-• quent pre- and post-election surveys dealt only with the Czech Republic. The < V.cch elections of 1996,2002, 2006 and 2010 have a small common set of ques-llons that facilitate exploring topics such as voter turnout, party choice, partisan-iliip, left-right orientation, trust in institutions and political efficacy.7 Each of ili.se lour post-election studies have been undertaken by CVVM using a slight modification of the omnibus quota sampling methodology to interview voters njjed 18 years or more (rather than the usual sample of respondents aged 15 years .....lore). While each of these surveys has been the subject of a number of pub- ln aiions exploring specific facets of individual elections, there have been rela-lively few publications examining trends across all elections (note, Linek et al. '(103; Lebeda et al. 2007). A central argument in a recent book entitled Zrazeni snu? [Betrayal of the l 'i ram | uses this set of post-election surveys, in addition to other relevant survey ■ I n.isets, to explain why Czech citizens' trust in politics declined and their sense ol dissatisfaction with party politics increased between 1996 and 2006 (Linek '010). This study argues that three key events: (a) the economic crisis of the late 1990s, (b) the party funding scandals of 1997-'98 and (c) the opposition agree-M>< ni of 2002 underpin Czech voters' sharp and permanent decline in political mil is faction with political actors and institutions. An exploration of the observed lunges in electoral participation between 1996 and 2010, again using post-elec-inhi surveys, concludes that Czech voter turnout is best explained using Valence I In oi y where the expected benefits of voting determines level of turnout. I he key implication of this research is that it is the 'supply' of parties; and 11| nee the nature and range of party choice available to Czech voters that de- ......i nes variation in turnout. All of the post-election surveys undertaken since I 'i'id contain a sufficiently large common set of standard questions that it is pos-ihle to make both cross-time and cross-national comparisons. This data harmo-iii .il ion work has been undertaken under the auspices of the Comparative Study 1)1 Electoral Systems (CSES) project; and has generated a considerable amount ol H-search where Czech political attitudes and behaviour form part of larger ii>ss-national analyses (Klingemann 2009, Dalton and Anderson 2011, Golder mil Slramski 2010).8 More will be said about CSES in a later sub-section of this Unfortunately, the unexpected nature of the Chamber Elections of 1998 mean that relatively I low (|uestions were asked in a post-election survey examining recalled electoral behaviour. I W Details of this comparative research programme are available at http://www.cses.org/. A |mi lal edition of Electoral Studies journal (Special Symposium: Public Support for Democracy: llimiilts from the Comparative Study of Electoral Systems Project, March 2008, 27(1), 581-776 .....(I by Ian McAllister) demonstrates the scope of research possible with this post-election •in vey data. 1110] [111] i neory, uata ana miaiysis chapter. Another topic that has been of particular significance to those interest ed in the survey based analysis of Czech electoral behaviour is the importance i >l class voting where data from IVVM and CVVM have been used (Matějů 19%, Vlachová and Řeháková 2007; Smith and Matějů 2011).9 Political parties have commissioned their own private pre-election polls ami used them for evaluating their election campaigns. For example, the Social Dem ocrats (ČSSD) are known from media reports to have received the results OJ weekly polling undertaken by STEM prior to the 2006 general election; and re ceived political marketing advice from the influential American polling firm, Penn, Schoen and Berland Associates that has strong experience in managing general election campaigns in the United States (Bill Clinton) and Britain (Tony Blair). Another point that is important to keep in mind is that the legal constraints on pre-election surveys in the Czech Republic are relatively stringent in compare tive terms, as the evidence presented in Table 3.1 reveals. The law imposing an embargo on the publishing of polls in the final week of a national election cam paign is similar to the legislation that existed in France between 1977 and 2002. In the French case, the law was changed because newspapers began to publish poll results during the embargo period outside of France in Belgium or on the Internet. To date, nothing similar has occurred in the Czech Republic.10 Pre-election surveys estimating likely party choices have a strong strategic im portance in the Czech Republic; and have generated considerable controversy. This is because the electoral system has a 5% threshold for representation in the Chamber of Deputies, and if surveys close to an election indicate a small party may fall below this threshold; then parties believe such estimates may lead to a bandwagon or snowball effect where support falls as voters decide not to 'waste' their vote on a party with no presence in parliament (Kreidl and Lebeda 2003; Lebeda, Krejčí and Leontiyeva 2004: 51-66; Lebeda and Krejčí 2007: 34-61)." The question of what constitutes accurate and reliable estimates of voter turnout and party support in pre-election (also known as 'trial heat') surveys is an important topic of ongoing research within the election surveying literature (note, 9 The Czech electoral system allows non-resident citizens to vote in person at the nearest Czech embassy or consulate. Using official election data that give details of migrants' party choices, Fidrmuc and Doyle (2006) reveal that migrants' preferences are different to their do mestic counterparts and reflect the values norms of their host country. 10 The embargo on pre-election polls is known to have been used in a strategic manner in French presidential elections (Baines, Worcester and Mortimer 2007). One of France's domestic intelligence services, Les Renseignements Généraux, is also known to have published mislead ing polls in an attempt to manipulate French public opinion (ESOMAR 1998: 2). 11 There have been anecdotal reports and accusations in the print media that survey based es timates of party support have been used in a strategic manner and without regard to the princi ples and ethics of scientific surveying espoused by professional market research organisations. Election Survey Research Mil. 3.1: International comparison of legal restrictions on election polling i tnintry ■ 1.11 alia ur.nia | Igium i ii niiiark i Ionia i inland ' ..ii nany India Ireland i itvia 11, norlands i Irw Zealand '.imth Africa ,wi!iien i imicd Kingdom i hilled States I i,nice iieece Poland i ill mania i .m.ida Portugal Romania Albania Russia ■ ,pain Bulgaria Cyprus Czech Republic Montenegro Slovakia Italy IVlll I uxembourg Restriction on publishing Pre-election polls Restriction on publishing Exit Polls None None None None None None None None None None None None None None None None 1 day (24 hours) 1 day 1 day (24 hours) 1 K days (30 hours) 2 days (48 hours) 2 days 2 days 5 days 5 days 5 days 7 days 7 days 7 days 7 days 14 days 15 days (Paracondicio principle) 15 days 28 days (1 month) None during campaign None, except for Victoria Yes None Yes Yes None Yes None Yes Yes None Yes None None Yes Yes Yes law makes no references to this form of polling. [1131 lt<>\ 3.1: Estimating likely voters in ;< pre-election survey Techniques for determining likely voters in an election have been developed in many political polling companies. The Gallup likely voter model is the most used method. The general approach adopted is; • If voter is not registered or says they won't vote, exclude as likely voter (leaves about 80.....90% ol respondents) • Seven questions are used to determine likely voter status among remaining respondents on the basis nl (a) Past behaviour: How often has the respondent voted in the past? (b| Practical knowledge of voting process: Do you know where your polling place is? • All respondents are given a score that ranges from zero to seven (0-7) Assumptions - Estimate the effective target voting population, i.e. proportion of eligible voters who will vote - Select a realistic voting turnout threshold or set of thresholds that reflect perhaps electoral participii tion in recent similar elections - Voters who are above the threshold are "likely voters" Step I: 1. How much have you thought about the upcoming elections for president, quite a lot or only a littlol (Quite a lot, or some = 1 point) 2. Do you happen to know where people who live in your neighbourhood go to vote? (Yes = 1 point) 3. Have you ever voted in your precinct or election district? (Yes = I point) 4. How often would you say you vote, always, nearly always, part of the time or seldom (Always, or neai ly always = 1 point) 5. Do you plan to vote in the Presidential Election this November? (Yes = I point) 6. In the last presidential election, did you vote for Al Gore or George Bush, or did things come up to keep you from voting?" (Voted = 1 point) 7. If "1" represents someone who will definitely not vote and "10" represents someone who definitely will vote, where on this scale would you place yourself? (Score of 7—10 = 1 point) Step 2: Adjust for not registered, say will not vote, and already voted Step 3: Adjust for the young as they are systematically less likely to turnout Step 4: Using demographic weights estimate profile of turnout on 0-7 scale Step 5: Estimate likely voter turnout at different thresholds and associated vote intentions Estimates of voter turnout using a Gallup poll, October 24 2004 Criteria for estimating turnout Registered voters - Unweighted Likely voters - Weighted Likely voters - 50.0% turnout Likely voters - 55.0% turnout Likely voters - 60.0% turnout Likely voters - 65.0% turnout Actual results - 56.7% turnout* George W. Bush (Republican) % John Kerry (Democrat) % 49-7 45.(, 49.0 46.2 52.5 43i, •"■4 43,5 51.9 44,| 51.6 44 (, _____ 50.7 48.3 7T^ TSUally h!gh ?n Nov- 2 2004- ie- 6-4% higher than in 2000, and the highest since Sg^ofZTe^Zt^ elCt7 made eS'timati°n °f Mkel> ™ters more difficult because,,,, mast prevfou:fe?5ons P " determmants ^Participation was different to those in eviden, ,„ Disadvantages of likely voter models I VotVmor V,°ter Tdel is.primari|y a motivation oriented explanation of turnout Vote motivation changes ,„ ways that are no. always directly related to voting behaviour [114] llm I Lilian 19X6, Martin, Traupotl and Kennedy .'005; (ichiian and king 1993; Aucncaux 2006). Box 3.1 demonstrates one strategy used by a major political mi eying company to obtain mote reliable estimates of electoral participation, iii.I hence patty support. Sucb tasks are fundamentally important in successfully predicting election outcomes. i / Senate Elections (1996-2010) nl ike all other levels of governance that have elected representatives, elections ■ ih'- Senate are unique in that there are never upper chamber contests across the '..... country (this only happened for the inaugural elections). A third of seats 11 up for election every two years. This staggered election schedule combined Hi, a majority run-off (two rounds) system, general lack of popularity and trust i ilir. institution has had a strong impact on the academic study of this type of • 11 Hid order elections (Lebeda, Malcova and Lacina 2009, Reif and Schmitt 1 111 In the first Senate Elections of 1996, a three wave panel study was under-'i' ,i by SC&C. Czech senate elections have two rounds, so the first wave of the Wild survey was fielded immediately prior to the first round (November 13-14); | >■ '' ond wave was undertaken immediately prior to the second round (Novem-#1 20-21); and the final wave involved interviewing respondents immediately ii' i Ihe second round.12 I his panel survey is invaluable because it is the only study in the Czech Re-hI'Im that facilitates exploring the dynamics of electoral participation; however, IiIn panel survey has been rarely examined taking advantage of this data struc-11'1 i in >te. Kreidl and Lebeda 2003). Czech senate elections are characterised by i niial voting suggesting that candidate characteristics influence vote choice. ■ ,'ll i '009) demonstrates using this panel dataset the importance of candidate II. i i'. on voting behaviour; and in addition reveals that the profile of the 'ideal mi 11'I.,ic" varies systematically across subgroups of voters. Hi' u- was also some senate election surveying undertaken on October 22-24 I lovember 3-4 1998 by Factum for TV NOVA; where questions focussed ii |',i dieting the results in each senate (single member) constituency by asking i• in on vote choice in the first round and vote intentions for the second round, mil Miters' perceptions of the senate campaign. Some post-senate election sur- iii" was undertaken by STEM in December 1996 and January 1997 using i In-, panel survey employed a probability stratified sample where 1,174 face-to-face inter-n wore undertaken with a representative sample of the electorate (18 years or more). For in iliilails see, http://sda.soc.cas.cz/data/0079/0079.htm (accessed 15/02/2012). [115] rhoory, Dutu jiihI Analysis many ol (his company's Ircncl scries ol questions (Toka ,'()()(): 143). None ol ilu data have been archived with ČSDA. To sum up, there are few examples of survey based analyses of electoral helim iour in Czech senate elections. However, there have been some studies using ek i toral and candidate data to explore the effects of the majority runoff system on piuiy representation (Lebeda 2011; Lebeda, Malcová and Lacina 2009; Lebeda, Vlachu w) and Řeháková 2009). Otherwise most of the survey data relating to the Czech Sen ate relate to (a) the public debate surrounding the establishment of an upper chain ber which was explored in IVVM surveys from late 1993, and (b) public trust in tin Senate and all other political institutions - questions that are asked frequently will) in the rVVM/CVVM monthly series of surveys (Herzmann and Rezkova 199 1) 3.3 European Elections (2004-2009) Since Czech accession to the European Union in 2004, there have been Iwtf Czech waves in the European Election Study series of post-election survey. " In addition, to a standard questionnaire designed to explore the motivations mm derpinning electoral participation in 'second-order' elections this comparalm research programme has also undertaken parallel studies of party manifesto! the European Election campaign and surveys of candidates standing for election This integrated approach to the study of supranational elections was especially evident in the EES (2009) study.14 To date there has been relatively little work published specifically on Europe an electoral behaviour in the Czech Republic. Analyses undertaken for the lira European elections in 2004 have focussed primarily on the actual election resnlH and exploring the spatial pattern of turnout and party choice (Linek 2004, I uu k and Lyons 2005, 2007ab). In general, much of the survey based literature mi voting in European Parliament elections across the European Union since I'W'J has focussed on testing the implications of the Second-Order-Election-Thesis n| Schmitt and Reif (1980) who argued that the lower salience of all contests ilim 13 The central goal of the European Election Studies (EES) is the comparative study ol nlm toral participation and voting behaviour in European Parliament elections. In addition, thomm such as the evolution of an EU political community and a European public sphere, citizens' pm ceptions of and preferences about the EU political regime, and evaluations of EU political pal formance have also been examined. For more details see, http://www.europeanelectionstuili.iH net/(accessed 15/02/2012). 14 A recent special issue of the Electoral Studies journal (Special Symposium: Electoral 1i" mocracy in the European Union, 30(1): 1-246, March 2011, edited by Sara B. Hobolt and Mink N Franklin) demonstrates some of the key topics examined in the 2009 European Election Study til 2009 that was funded and managed by the PIREDEU project funded by the EU. For details iitm, http://www.piredeu.eu/ (accessed 15/02/2012). tiection Mnvi'y Hesearcn ii, a r, i in a I c led ions is ly pitied by lower turnout and higher support lor small II Incumbent parties. Survey data on European Elections in the Czech Repub-If limy be uccessed from the EES website. • 't, .,.onal and Local Elections (2000-2010) .....il tier of government was created in the Czech Republic in the late HpNbi and the first regional elections were held in 2000. To dale, there have >■ 11 u. >• rounds of regional elections in 2000, 2004 and 2008. Most of the re-■4lvli on ihcsc regional contests has not used survey data to construct for exam-■ I i'ii i'li i a I level models of turnout or party choice; and this remains an area uiimily for future research work (see, Vajdová 2001, Balík, Kylousek, I..11.1 .-t 2005, Šaradín 2008, Eibl et al. 2009, Kostelecký 2007). A num-i questions were included in a CVVM survey of November 2008 that fo-i mi exploring differences in individuals' attitudes toward electoral partic-Iji.Hi.Mi ľ. i ween different types of elections. One of the central motivations for ^Bjl hwarch is the Second-Order-Election-Thesis (noted earlier) which argues Hp Void-, (and parties) have a hierarchical view7 of elections.15 Mi n decisions relating to turnout and party choice are characterised by (1) Batumi is lower than in national elections; and (2) voters are more likely to sup- E**! umal I protest or peripheral parties rather than express a 'normal vote' for one |n< mainstream parties, as would happen in a typical general election. Con-iii 1111 v. regional elections are used by politically engaged citizens to express i ľ m rs about the performance of the incumbent government, often by ex-IxlllK dissatisfaction (Reif and Schmitt 1980; Reif, Schmitt and Norris 1997). •m..... b these elections are 'non-salient', they are important in the signals they ^Ěm\ i, i. m incumbent governments. For example, the 2004 European elections in HbCVi-cb Republic resulted in the resignation of Prime Minister Vladimír Špidla mul iln 11 n iporary collapse of the Social Democrat led government due to a poor 1 .....by the ČSSD in these 'mid-term' elections. i......portant at this point to note that political attitudes within the Czech Re- Hptili ii the regional level have been examined using mass survey techniques, n , uiiples of about a thousand respondents have been used to explore dis-llti .....ional cultures (Vajdová and Kostelecký 1997; Kostelecký 2001; Kostě- ní mder elections are defined as general (lower chamber) elections that yield govern- i.....nlrast, second-order elections relate to the election of sub-national assemblies, na- aiii iis such as the president who do not perform a strong executive role or national refer-in nil lopics such as the European Union. [116] [117] I llľoi y. Ulllíl .111(1 /VIUIiyM*. lecký anil Čermák 2004). Such research slems Írom Ihc apparcnl slahilily ol |>ai ty choice across all elections in the Czech Republic since 1920: a fact evidenl In electoral maps (Jehlička and Sýkora 1991; Kostelecký IW3, 1994; Kosleleckv Jehlička, Sýkora 1993; Voda 2011). Such examinations of the spatial basis Ibi differences in political attitudes and electoral behaviour also includes related themes such as the Czech public's sense of local, regional and national identít) (Nedomová and Kostelecký 1997; Vlachová and Řeháková 2004, 2009). The use of mass political surveys for the study of local elections in the Czech Republic is very limited for three main reasons. First, local contests relate to fl level of governance that has limited powers; and are thus considered relatively unimportant. Second, local elections exhibit strong candidate effects implyinj that the results from these contests do not provide a strong indication of publli satisfaction with government performance. Third, as local elections are domi nated by a multitude of local issues and personalities these characteristics an not well suited to examination through nationally representative sample survey, However, it is important to note that CVVM have asked questions about turnout and a small number of other topics in surveys fielded in 1990, 1998, 2002, 20l)fi and 2010. To date, there have been no published survey based analyses of voting in local elections. Local election scholarship in the Czech Republic has tended to use ol ficial rather than sampling data. For example, Kostelecký (1996) examined how local elections acted as the foundations for the establishment of local elites. ()ih erwise, local government has been examined from a historical or institutional perspective. The most notable exception to this evaluation is a large survey pro ject undertaken for the local elections of November 1994 by STEM. With a rol ativcly large sample (N=l 1,672) this pre-election survey, conducted in late M;iy and early June, explored voting behaviour in terms of left-right orientation ami local political issues; where the goal was to map out regional differences in no litical attitudes. This data has not been archived (Tóka 2000: 127).16 3.5 Exit Poll Survey Data (1990-2010) One of the most important types of election based surveys are those underlay en directly outside polling stations on Election Day in all lower chamber ek tions since 1990. In addition, there have been exit polls for the EU accession n erendum and the two European Parliament elections. Many of these exit poll 16 It seems this data was used by Jan Hartl, head of STEM, to write a series of newspaper mil cles that provided a profile of contemporary Czech political parties. t laction Suivoy Rosttdrcl [118] i Uvn commissioned hy ( V.cch Television ami TV NOVA for (heir immedi-H iin i election coverage where the goals have been (a) lo predict the outcome wMoro all the counting of ballots has been completed, and (b) to provide some 1 * i1 explanations of the election outcome. 'typically an exit poll questionnaire asks: party choice in the current and previ-tilln elections, timing of voting decision, party choice in the last general election, rtial perhaps some items on perceptions of parties, party leaders and priorities for thf next government.17 All of these exit polls have large sample sizes where re-HMil S< 'SiC exit polls have about 15,000 respondents. Here is a brief overview of ill exit polls undertaken in the Czech Republic. * federal Elections 1990: Gallup International, INFAS (Germany) and IVVM * Ivdcral Elections 1992: INFAS (Germany), IVVM and FACTUM-non I .iluda • ('hamber Elections 1996: (i) IFES (Austria) and SC&C for Czech Television; (ii) INFAS (Germany), Sofres-Factum for TV NOVA18 • < hamber Elections 1998: (i) IFES (Austria) and SC&C for Czech Televi-moii; (ii) INFAS (Germany), Sofres-Factum for TV NOVA i hamber Elections 2002: SC&C for Czech Television I niopcan Parliament Elections 2004: SC&C for Czech Television « < hamber Elections 2006: SC&C for Czech Television ■ < lumber Elections 2010: SC&C and SPSS CR for Czech Television19 If) mline occasions exit poll surveys have formed part of an election study series, tlfc was (he case in May-June 1996; when STEM undertook two large pre-elec-H polls (N=6,205 and 5,455) and an exit poll (N=8,846). Many of the items in M's I rend series of political questions were fielded in these surveys (Toka DUO; 138). The Exit Polls for 1992 and 1996 are freely available from the Czech Science Data Archive (CSDA). However, the results for more recent exit " lie not publicly available although it has been possible to purchase cross-i-.1-111.11 ion (ables of this data from SC&C. 61 &C report on the exit poll of 2010 demonstrating the scope and use of such data is Hilliln nt: img2.ct24.cz/multimedia/documents/17/1699/169810.doc (accessed 15/02/2012). l.-l-.i (7000: 138) indicates that STEM also undertook an exit poll for the Chamber Elections ' i-iii I lowever, it seems more likely that this was a standard post-election cross-sectional "......its associated with exit polling are relatively high and prohibitive for most media or- .......H was reported that SC&C were paid 2.4 million Kcs and SPSS CR, 2.2 million Kcs -Minutely for the 2010 exit poll commissioned by Czech Television. ■Hi- www.louc.cz/10/2210601.html (accessed 15/02/2012). [119] Lixil poll daki arc typically used by academics Id build piohlcs ol Ihe voters for different parties. Linek and Lyons (2()07a,b) have used exil poll data to makP estimates of party switching behaviour across pairs of consecutive elections and compare the results with (a) other post-election surveys (e.g. the Czech wave ol CSES) and (b) ecological inference statistical estimates using official election re suits. At this point, it makes sense to give some recent practical examples of Ihe insights that may be gained from examining exit poll survey data. 3.5.1 Some insights from the SC&C Exit poll (2010) There are a number of key questions that are particularly well suited for study with an exit poll with a large sample that records respondents reported vole choices within a few minutes of casting a ballot. Consequently, many of Ihe methodological problems associated with eliciting recalled vote choices in post election polls such as voter over-reporting (i.e. incorrectly claiming to have vol ed in an election) and providing inaccurate accounts of party choice because o I social desirability and other survey response effects are minimised. One of the important questions that may be addressed in an exit poll with a very large sain pie (N>10,000) are the structural bases of party choice. The Chamber Elections of May 28-29 2010 were one of the most dramai ic over the last two decades. There were four key trends. First, the two largest parties lost close to one and a half million votes when compared to their performance in 2006. Second, two parties lost all their representation in parliament (the Christian Democrats, KDU-ČSL and the Greens, SZ). Third, two parties losl their leaders on the basis of a poorer than expected electoral showing. Fourlh. two new centre-right parties made a breakthrough with TOP 09 (Tradition, Re sponsibility, Prosperity 2009) and VV (Public Affairs) becoming the third and fifth largest parties in parliament respectively. In contrast to previous elections, the parties of the right won a convincing victory winning 118 out of 200 seals. In sum, the official election results suggested a significant change in the nature of party competition. These data suggest that some of the larger and more established parties losl significant levels of support between 2006 and 2010. An examination of vole switching between these chamber elections using estimates from the SC&C Exil Poll (2010) shown in Table 3.2 reveals that KSČM has the most loyal voten (82%) and both the Social and Civic Democrats lost significant amounts of sup port (ČSSD: 35%, ODS: 49%). Although the Christian Democrats had a highei loyalty rate (60%) than ČSSD and ODS, their loss of support mainly to TOP 09, ODS and VV resulted in their failure to exceed the 5% threshold to enter pari i a ment. [120] l ■ i■ 11- t./: Vote switching in the chui nliľi e l'ClKllľ. hľlween 2006 .ind 2010, percent li,-t . II, ;l i'.nly ch i/a; in 2006 I'mty/yoar »* čssd ods l< s f: m kdu čsl sz Other Did not No right Total vote to vote ČSSD 65 3 6 4 5 5 14 8 22 ODS 2 51 1 6 9 7 13 20 21 ^ KSCM 6 1 82 2 3 2 8 1 11 Mill čsl 1 1 60 3 3 1 2 4 fl SZ 1 1 1 19 3 3 3 2 j" Other 2 3 2 2 6 18 10 12 5 I op 09 4 26 1 16 30 20 21 30 17 K W 8 10 2 5 15 23 19 17 11 :;i'oz 6 2 1 2 5 7 7 5 4 Suverenita 4 2 2 2 4 11 5 2 3 lol.-il 100 100 100 100 100 100 100 100 100 $mmw: Exit Poll 2010, SC&C and SPSS Czech Republic for Czech Television. Surveying was undertaken M.iV ?H 29 2010. Total sample size was 25,380 respondents interviewed in 370 districts. (Ote ili.it column totals, i.e. source of party support in 2010, sum to one hundred percent. The final col- "iii.....i Ihe right indicates the total level of party support in 2010 excluding non-voters: who are by defi- Mllhin not interviewed in exit poll surveys. The bold numbers on the diagonal indicate levels of consist-mhI m loyal voting in the 2006 and 2010 elections. This table should be interpreted as follows. Almost (wn thirds (65%) of those who voted for ČSSD in 2006 also voted for this party in 2010. The remaining W% Bwitched their votes away from the Social Democrats and voted for rival parties such as KSČM |«%), I OP 09 (4%), VV (8%) and SPOZ (6%). I lie Green Party's limited electoral appeal and difficulties in maintaining par- .....11 y while in coalition with ODS and KDU-ČSL meant it has had represen- i u ion in the Lower Chamber for just one legislative term (2006-2010) since the i1 Miy was founded in 1990. The estimates in Table 3.2 indicate that most green iJUty support drifted to TOP 09 (30%), VV (15%) and ODS (6%). The voter || m it ion estimates presented in Table 3.2 demonstrate that the success of new i ui.ies such as TOP 09 and VV was based on two key mechanisms: (a) vote witching by those who reported voting for ODS and ČSSD (but not KSČM) in 'inid, and (b) attracting first time voters and abstainers in 2006. < )ne of the key themes in the post-election commentary was that the declining i n i unes of the established parties and emergence of new parties had a strong age ......ponent. The exit poll profile of party support for all parties is shown in part . not Table 3.3 reveals that support for some established parties (ČSSD, KSČM nuI K DU-ČSL) was more concentrated among the older cohorts (40 years plus). In t on trast, the new parties (TOP 09 and W) attracted much higher than aver-iif>r levels of support among three youngest cohorts. Notwithstanding these age [121] Table 3.3: Exit poll estimates of age and party choice for the Chamhei I lections of 2010 (a) Composition of specific party choice by age cohort Age cohorts Party 18-19 yrs 20-21 yrs 22-29 yrs 30-44 yrs 45 59 yrs 60 yrs+ Total ČSSD 2 2 7 22 29 39 101 ODS 3 3 15 33 25 20 99 KSČM 1 2 6 18 35 39 101 TOP 09 5 6 20 37 21 11 100 vv 5 5 18 36 23 14 101 KDU-ČSL 2 2 12 27 28 29 100 SZ 5 5 26 38 17 9 100 SPOZ 3 5 19 36 24 13 100 Suverenita 1 3 10 27 32 27 100 (b) Vote choice by age cohort Age cohorts Party 75-79 yrs 20-27 yrs 22-29 yrs 30-44 yrs 45-59 yrs 60 yrs+ ČSSD 13 10 11 16 25 37 ODS 19 19 23 23 20 17 KSČM 2 6 4 7 15 18 TOP 09 28 28 24 21 13 8 VV 15 14 14 13 10 6 KDU-ČSL 3 3 4 4 4 5 SZ 4 3 4 3 2 1 SPOZ 4 5 6 5 4 2 Suverenita 1 3 2 3 4 4 Other 11 9 8 5 4 2 Total 700 700 700 700 707 700 lloclion Survey Himonrch Source: Exit Poll 2010, SC&C and SPSS CR for Czech Television Note that total percentages in both tables sum to one hundred subject to rounding error. The data in table (a) should be interpreted as follows. Popular support for CSSD is mainly composed of older voters, i.e. 29% of 45-59 year olds and 39% of 60 years or more. In contrast, in table (b) a plurality of 18-19 year olds (28%) voted for TOP 09, while a further 19% voted for ODS, 15% supported VV and 13% CSSD. differences, the 'middle-aged' (3GM4 years) constituted the most important demographic for many established (ODS, SZ) and new parties (TOP 09, VV and SPOZ). This constellation of parties suggests that this middle aged group constitutes a demographic heartland for right or centre-right wing parties.20 20 The age profile of voters for the centre-right TOP 09 and SZ are most similar being concentrated among those aged 18 to 29 years. In the cases of the leftist ČSSD and KSČM these two [122] I In- parly support by age cohort estimates presented in part (b) of Table M facilitates viewing the most popular parlies within each cohort. These data show Hi ii I OP 09 was generally the most popular among young voters. In contrast, nlilcr voters' preferred party in the Chamber Elections of 2010 was the Social Democrats (ČSSD). The general implication from Table 3.2 is that there is a hrnnd partisan division within the Czech electorate based on age where (I) oldei ni i ns socialised under the communist regime are more social democratic than I i the younger post-communist generation who are generally centre-right, and i I) a middle aged group who came of age around the fall of communist who ex-i ubil a right wing (ODS) orientation. Hiving examined "snapshot" (cross-sectional) surveys in the previous sec lions, it is now time to turn our attention to survey data that facilitate exploring thi dynamics of attitude change at the individual level with a panel survey do i (i Panel Survey Data on Political Topics Most surveys are cross-sectional in that the respondents are interviewed on one occasion. These 'one shot' surveys provide a picture or snapshot of society at a pceific point in time. However, such surveys do not facilitate studying directly 11 it process of social or political change. In order, to examine change at the individual level using a representative sample of the adult population; it is necessary |0 interview the same respondents at two or more time points. This is the basic logic behind panel surveys. Use of repeated cross-sectional panel surveys (i.e. Nume questions but different respondents in each poll) may be used to indirectly infer attitudinal change, but they are less suited to this task than panel survey ■ liia (Vinopal 2009b). It should be noted that there are important panel surveys undertaken in the Czech Republic such as EU-SILC and SHARE that are useful i or examining some political questions; however, the central focus of these research programmes is on social inequality and ageing respectively, and not pol-ilics.21 11.iities are most similar because most of their voters are aged 60 years or more. The implica-lion here is that with demographic metabolism rightist parties will attract more support in future elections as left wing voters exit the electorate through mortality. 21 More details on these panel surveys may be consulted in other chapters of this book and at the following websites: European Union Statistics on Income and Living Conditions (EU-SILC) -hi i p://epp.eurostat.ee.europa.eu/portal/ page/portal/microdata/eu_silc !.urvey on Health and Aging in Europe (SHARE) - http://www.share-project.org/ [123] Modlou Siirvny RoHOlirch There have been less than a handful of panel suiveys dealing explicitly with political themes in the Czech Republic. As noted earlier, there was a two wave panel for the first senate elections in 1996 thai allows one to study electoral pai ticipation and candidate choice where the attitudes before and after the election may be compared. Another panel survey with political questions was a four wave study undertaken in the town of Klatovy (an urban administrative centre in tin Plzeňsky kraj / Pilsen region, which contains about 23,000 inhabitants) between September 1999 and November 2000. This unique study taking inspiration from Lazarsfeld et al.'s (1944: 155) work on the link between interpersonal communication and formation of political at titudes via a two-step flow of information model was led by Prof. PhDr. Hynek Jeřábek CSc. This political attitudes survey had an initially large sample of ovei 2,000 respondents that eventually declined to a final panel size of about 500 re spondents due to well-known panel attrition effects. This study explored the stn bility of local citizens' political attitudes and values and contains standard items for measuring left-right orientation, etc. (note, Jeřábek 1999; Schubert 2010). lit date this panel survey has not been archived with ČSDA. A more ambitious panel study project examining political attitudes and me dia agenda-setting was implemented by CVVM over a twelve week period from April to July 2008. A panel of about 650 respondents undertook on a weekly ha sis to send a self-completed questionnaire to CVVM. Using a postal mode of in terviewing in a panel survey is unusual as much panel surveying is currently tin dertaken via the Internet. An examination of the dynamics of Czech citizens' attachment to parties (party identification) revealed that the social-psycholoj'i cal or social identity basis for stable party support, seen by many scholars as | key foundation for a stable democracy, is strong. However, the number of eitl zens with some sense of party attachment constitutes only a minority of the I o tal electorate (Linek and Lyons 2009). This panel survey has not been archived with CSDA. 3.7 Inter-election Political Opinion Polling Political opinion polling is undertaken frequently during inter-election periods where media outlets, parties and interest groups of various types commission surveys to examine specific topics. Most of this commercial polling is undci taken by a handful of companies such as STEM, SC&C and Factum invenio 22 According to Toka (2000) a number of polling companies such as Factum and STEM incliul ed a standard set of political question in their omnibus monthly polls throughout the 19!ltit ti" iij'c.regalecl results from such research are often published in (lie print me ' li i possible with this data, for example, to compare different polling com inmliV estimates of likely party support in the next elections.23 However, the in-illviilii.il level survey datasets are most often unavailable; and they are currently ' u. hived in a systematic manner with the Czech Social Science Data Archive (CSHA). In this respect, researchers need to make representations to a polling i N|tlH'y regarding specific survey datasets. I'm innately, the situation is different with CVVM as this is not a commcr- ■ il m.n kel research organisation. Its primary purpose is to undertake surveys i 1 eeh citizens' attitudes as a public service, and many of its monthly surveys "i mi two sections: (1) a standard battery of items that are asked in all surveys l it least periodically [see below], and (2) special modules commissioned by ■ id. mic researchers examining specific topics such as public attitudes towards ...... ■ participation in politics. All of these monthly surveys are archived with ili. 1 /eeh Social Science Data Archive and are freely available for analysis by .....h. is 11" runge of political topics that have been the subject of CVVM surveys is ..... mi I almost all topics of public debate have been examined on least one oc- ■ ..il I in fortunately, there is as yet no searchable 'question bank' as provided He Ihe websites of the German and Norwegian Social Data Archives that would ■lliiw a researcher to identify which surveys examined specific topics. Nonethe-Hh*. exploration of the archive of CVVM press releases and its bi-annual maga-Bjflt* NaXe společnost is possible through a 'search' feature on the CVVM web-" thereby identifying questions and surveys of interest.25 11 v\as noted above that CVVM asks a standard battery of questions each Hhtiuli and an additional set of questions periodically. The standard set of ques-HlMlft nsked in all polls (beyond the socio-demographic items) is intention to par-Hblpiile in elections, vote intention, closeness to a political party, left-right ori-..i in.hi, satisfaction with the political situation and trust in political institutions li. i lie President, government and houses of parliament. Ideally, there would . ■ .mibined individual level data file containing all the standard questions nli i 11.unionised set of socio-demographic variables. Unfortunately, such an It "in il level repeated cross-sectional dataset does not currently exist. This is i.ni.' ol this data has been archived with ČSDA. Iff fin iixtimple, at the STEM website (http://www.stem.cz) there are monthly estimates for "i. mi ions that go back a number of years. There appears to be no single web page that "i ill this monthly vote intention data in a spreadsheet format allowing the plotting of Rftlli en more detailed statistical analysis. ' 1 .....iverview of the main political survey variables from 1990 to 1996, see Toká (2000: ■i ww.cvvm.cas.cz/index.php?lang=2&disp=vyhledavani (accessed 15/02/2012). [124] [125] 1 henry. Data and Analysis because there arc considerable problems in harmonising questions and response options that have changed over the last two decades; and this is especially lnic lor surveys from the early 1990s. The opportunities offered by the construction of such datasets are evident in recent work by Lukáš Linek (2010), which employs cohort analysis to examine citizen support for the Czech Communist Party (KSČM); where it is argued using CVVM data that political socialisation is a key determinant of long-term support for this party. In this respect, some commentators' prediction that KSČM would disappear within a short period proved to be incorrect. According to Linek's (2008b) estimations this party will have sufficient popular support to remain in parliament until the early 2020s, and possibly beyond. In short, to paraphrase one of Oscar Wilde's more famous epigrams the imminent death of KSČM has been greatly exaggerated. One of the most important political events since the Velvet Revolution was the dissolution of Czechoslovak Federal Republic in 1993. In comparative terms, this event is important because it represents one of the few examples of a peaceful dissolution of a federal state. Typically, federal states disintegrate with considerable violence as happened in Yugoslavia during the early 1990s. For this reason, survey data on Czech and Slovak political attitudes is very important because it provides invaluable information on the citizen or mass basis for the failure of the Czechoslovak state.26 In this respect, an AISA survey of political attitudes in May-June 1991 with hour long face-to-face interviews with 1,260 respondents provides an important opportunity to explore attitudinal and value differences that might have underpinned dissolution (see, Rose 1992). Much of the literature on political attitudes in Czechoslovakia under communism stresses the importance of the 'national question' in key historical events such as the Prague Spring 1968 and the fall of communism in 1989 (note, Dean 1973; Steiner 1973; Kusý 1997; Hilde 1999; Brown 2008). It is important to conclude this sub-section on inter-election survey data with an example of research on political attitudes where the goal has been to explore opinion change across time within the Czech Republic. One of the earliest political attitudes surveys for which there are individual level data is a study entitled Postoj občanů k politice (Attitudes of Citizens towards Politics) which was undertaken in May 1968. The fieldwork for this survey was fielded by ÚVVM (a predecessor to CVVM) and the goal of this research was to provide data for the 26 Contemporary IVVM surveys indicated that there was not majority public support for the dissolution of the Czechoslovak federal state suggesting that this was an elite led decision (Young 1994: 11-18; Kraus 2000; Deegan Kraus 2000: 254-256). A CVVM survey fielded in December 2007 revealed that a plurality of Czechs (47%) thought the breakup was unnecessary, 30% believed it was, and the remainder (23%) had no opinion. [126] Election Survey Research I'iague Spring political reform progra.....ie (see, Brokl el al. 1999; Lyons 2009). Mosl of the questions in the May 190s survey were replicated forty years later in May 2008 where the goal was to sec if Czech citizens' democratic attitudes and values were significantly different under communism (1968) and liberal democracy (2008). The results of this research presented in Lyons (2009) reveal that there is a remarkable stability in attitudes across time where Czechs living under communism had very similar attitudes to their descendents living in a multiparty liberal democracy. This finding is important because it suggests that democratic values can exist independent of prevailing political institutions; and the idea that Czechs had to iearn democracy' in tabula rasa manner in the 1990s is an over-simplifi-■ alion of a more complex political reality.27 t.8 Examples of inter-election dynamics < >ne of the most important features of inter-election periods is the evolution in support for political parties in regular opinion polls undertaken by CVVM, STEM and Factum Invenio that are regularly reported in the media. In addition, parlies also pay great attention to their performance in local, regional, senate and European elections as these events are seen to provide important informa-lion about the popularity of a party in the next general election. Within European political science there has been considerable research on differences in voter participation and party choice across consecutive general and European elections. -L8.1 Evolution of electoral preferences To keep matters simple, the estimates of vote intentions presented in Figure 3.2 locus on a single year - 2004; and the first European Parliament elections held in the Czech Republic on June 11-12 2004. Competition for the 24 seats during the campaign was primarily candidate-centred. A rather lacklustre and lukewarm campaign was dominated by a curious mix of candidates: the first and only ('zechoslovak cosmonaut, Vladimir Remek, who went into space on board Soy-ii/ 28 in March 1978 (KSČM); German-based but Prague-born porn star Dolly Buster or Nora Baumbergerová nee Dvořáková (NEI, Independent Erotic In- 27 An interesting comparative analysis of voters and politicians learning democratic politics In the decade after the fall of communism in the Czech Republic, Hungary and Poland is given Iworzecki (2002). For a comparative analysis of economic voting in Central and Eastern Europe, mm; Pacek (1994), Fidrmuc (2000), Fidrmuc and Doyle (2003), Tucker (2006), Roberts (2008), Lew-r. Beck and Stegmaier (2008) and Fauvelle-Aymar and Stegmaier (2008). [127] Theory, Data and Analysis Election Survey Research Figure 3.2: Monthly trends in vote intentions for elections to the Chamber of Deputies during 2004, per cent 20%-\ ..-cssd -^ktu me: .DE -"-us —-wtatnl -•-kscm ---sz . dk.no p.Mty 10"o- Jan Feb Mar Apr —r~ Jun ~l— Jul Sep Oct —i— I lov i MM Source: CVVM omnibus surveys, 2004 Note that the monthly estimates of party choice among those fairly or very likely to vote are ba samples of approximately one thousand respondents and the confidence intervals on the vote intent estimates are ±3%. Ostatni refers to small other parties. itiative); former general director of TV NOVA, Vladimir Železný (Independent Democrats); and Viktor Kožený (OFD, Citizens' Federal Democracy) an cntn preneur later charged with embezzlement on a massive scale during the votu h er privatisation of the 1990s. Pre-election polls undertaken by CVVM indi< nl ed that ODS would secure 26% of the vote followed by KSČM (12%), CSS I» (10%), KDU-ČSL (8%) and US-DEU (<5%). In this election there was a record low turnout of 28%, although CVVM •• pre-election poll in May had predicted a participation rate of 63%.^ As predii i 28 This is a good example of the problems over and miss lopoilina onr.ountomd in iimiiii pm electoral surveys to predict voter turnout and party suppnil Sim linx 3.1. Moro will ho s.inl mi this topic in chapter 7. 11 ODS did well winning 30% of the vote (9 seats) followed by KSČM {20%), SNK (11%), KDU-ČSL (10%), ČSSD (9%), Independent Democrats (8%) and (iieens (3%). As the ruling Social Democrat vote collapsed to a third of what it been in the previous general election in 2002, ČSSD leader Vladimír Špidla Wits forced to resign and there was a government reshuffle. \n examination of the vote intention data for 2004, shown in Figure 3.2 re-Vťiils lour key patterns. First, there was a rapid growth from 5 to 25% in sup-for 'other' (ostatní) parties on the eve of the European elections. Popular npport for these small other parties declined rapidly after the European Parlia-llii ni elections, although it resurged somewhat later in the year. Second, there WMN a doubling in support for the Green Party (SZ) across 2004 from about 6 to i 1 . indicating some popular basis for its breakthrough in the subsequent 2006 1 ii uiiber Elections. Third, there was considerable volatility in support for small ni n-s such as Union of Freedom (US) and to a lesser degree with the Independ-| Dice parly (NEZ). Lastly, all of the main parties retained largely constant levels ni'port across the entire year. Overall, the main message evident in the internou dynamics presented in Figure 3.2 is one of complex short-term chang-I |n thai have their origins in real opinion changes and methodological features of I llliveying such as sampling error (±3%). mi.' Public trust in politics and economic sentiment 1..... election surveys also ask a variety of questions regarding citizens' attitudes 1»iwin (I the political regime and institutions of representation. Within political Ii in ' many scholars argue that there is a qualitative difference between trust in ni millions and attitudes toward office holders. There is reason to doubt that ii i'undents participating in inter-election surveys do in fact separate the perforin ni' i of institutions from office holders when making responses: as the logic of lln misi item in political attitudes surveys assumes. 11 li <' V VM time series data presented in Figure 3.3 is composed of three dis-■bk'l series: (1) satisfaction with the regime; (2) satisfaction with national institu-"ii "I political representation; and (3) consumer confidence. The main pattern nl' ni in figure 3.3 is that all six series are correlated, where the rise and fall l ili' public mood' is evident across all survey indicators. It is not possible to Ii Inuiively slate without a more detailed time series analysis such as Vector Au-HpfpjtivsNion (VAR) the direction of causality. An example of such an analysis is ■ ' ' n in llie next sub-section. Another important feature of Figure 3.3 is that sat- ■ i ii I ion with the regime (or political situation) appears to be a composite meas-nii "I iiusl in political institutions. These questions are reasonably strongly cor- Figure 3.3: Trends in trust in government .mil parliament .mil < on.mini •.enlimenl in the Czech Republic. 1996-2006 (quarterly) -25- Satisfaction with political situation —~~~ President '—*"* Government ( \ i. iiiil>< i ol I ><-puties ~ Senate Consumer confidence o o o, a o, a d, a, 6, a a <3 6 a a a a 8 a 8 a 8 SSoSoooSSSSooo oooooooooooooo cncnc\jcncncn](^icnc\jc\|c\1c\jc\icn cocdr--r-.coa>05cj) cdcr>05oj05cdct>a> Pairwise correlations Satisfaction Trust in Trust in Trust in Trust in Consumer (Bonferonisignificance) in politics President Government Chamber Senate confidence Satisfaction with the 1.000 political situation Trust in the President 0.213 1.000 1.000 Trust in the Government .798 .280 1.000 <.001 1.000 Trust in the Chamber of Deputies .719 .091 0.670 1.000 <.001 1.000 <.001 Trust in the Senate .642 .284 .528 .658 1.000 <.001 1.000 .010 <.001 Net consumer .488 .294 .499 .498 .768 1.000 confidence (Eurostat) .016 .878 .012 .012 <.001 Sources: CVVM omnibus surveys, 1996-2006; Eurostat economic confidence surveys, 1996-2006 Note that the level of trust for the main political institutions and is taken from the responses of those aged 18 years or more between 1996 and 2006. This data has been aggregated to quarters and represents between 886 and 4,683 responses. The consumer sentiment time series is based on Eurostat's consumer confidence survey undertaken monthly in all EU member states with national samples of one thousand respondents. The consumer sentiment estimates are seasonally adjusted and represent the balance between positive and negative responses, and for the most part during this time period were negative. [130] n l.ileil (ranging from .(>•! lo MM . 11; •; ■ i ■. 1111; ■ that I here is allimdinal linkage.'"The • •iiM-.tenlly higher level ol publii support given lo the President suggests that *• >eli citizens have greater Irusl in non partisan institutions.'0 I he period under consideration is important because it includes a phase of pi'onomic decline between 1997 and 1999, which had its origins in a currency in.I hanking crisis accompanied by a number of political scandals. This led to in unscheduled general election in June 1998; and a series of austerity packages ili.it rapidly cut public spending. The polling data on the left of Figure 3.3 shows ill ii economic and political turmoil was accompanied by a decline in consumer NCntiment, trust in political institutions, and satisfaction with the regime. In general, the correlation between consumer sentiment and the political indicators i uij'cs between .34 and .49 suggesting a moderately strong relationship. It is necessary at this point to stress that great care is required when interpreting correlations of time series data. Strong bivariate correlations may be spurious iih apluring little more than common trends due to third factors such as partisan-hip in the political trust variables trends, or may be due to chance. Therefore, 11 is not valid to infer causality from the correlations reported here without un-ili flaking appropriate time series econometric modelling - a topic for further re-■ arch. II le strongest correlation observed is between consumer confidence and trust in i he Senate (r=.78). This is a surprising and puzzling relationship. If this correlation is not spurious, one possibility is that trust in Senate is more strongly asso-> ialcd with citizens' personal resources, such as higher levels of education, political knowledge and income as is evident in other research. And it is this subset of Mi/ens who are more sensitive to changing economic sentiment because many members of this group are key figures in business. For the moment such explanations must remain speculative and represent an important avenue for future re- - arch, as little has been written on how economic factors shape political satis-I act ion ratings in the Czech Republic. The CVVM time series data presented in Figure 3.3 demonstrate a number of important lessons when working with inter-election survey results. First, the responses to sets of political attitudes questions may be correlated indicating the presence of a more general public mood. Second, the manner in which respond-• nIs answer questions may not always reflect the logic of the question design. I lere we see that changing levels of trust in institutions appears to be driven by "i An analogous pattern is evident in individual level analyses of ISSP data (see, Linek 2010: se 59). :n A similar phenomenon was evident in the higher levels of government satisfaction given to the lechnocratic government of Jan Fisher which was in office between May 8 2009 and June 25 2010. [131] ihľoiy. il.il.i iiid aniilysis Lhc perľormancc oľoľľicc holders. Third, llie economic climate is also important where changing levels of consumer sentiment is matched by variation in polili* n attitudes. Fourth, establishing a causal relationship between time series variable*, requires careful modelling in order to avoid making invalid inferences because o| failure to take into account factors such as spurious correlation, autocorrclalioi trends and seasonal variation. 3.8.3 Causal links between public mood and trust in political institutions One of the salient features of Figure 3.3 is the similarity in the trends obsei vi ij between satisfaction in the political situation and trust in the Government, ('bam ber of Deputies, the Senate and to a lesser degree the President. Public satisliti tion with the current political situation would seem from Figure 3.3 to be an m dicator of the 'public mood' reflecting Czech citizens' general evaluation ol ill politics/1 One obvious question to ask of the trends observed in Figure 3. \ m what is causing what? For example, does public trust in the various political m stitutions determine the overall public mood? Or perhaps, it is the public hmhhI that is shaping the level of trust in the President, Government and the House, ol Parliament? Alternatively, the situation may be more complex where trust m out political institution determines trust in another resulting in a complex set ol >li rect and indirect relationships. Given the possibility of complex relations between the time series variable! shown in Figure 3.4, it makes sense to construct causal models that allow all o| the trends to be interrelated. One statistical method of simultaneously treat my all variables as both causes and consequences of each other is to estimate a Vi i tor Autoregression (VAR) model. In order to keep matters simple and to ensiiii that the model estimates are stable, the VAR model estimated will be a pars i i tin nious one containing four political variables where trust in the Senate is noi i sidered.32 The essential logic of this model is best shown with an example. Trust in go\ ernment at time 2 is said to be determined by trust in the Government ai m m 1 plus trust in the Lower Chamber at time 1 plus satisfaction in the political smut tion at time 1 plus trust in the President at time 1.33 This same logic applies to I III 31 This mood concept is similar to Stimson's (1999, 2004) 'public mood' or 'policy n......I m the sense that it refers to a general orientation toward politics that changes systematically • >vfl| time. However, the public mood here is different in that it is based on a single item rathoi llmii n composite set of measures subject to a time series factor analysis. 32 This strategy is followed for two reasons. First, there is a shorter time series for the : .minis as it did not come into existence until late 1997. Second, additional analysis reveals thai tnial m the Senate is independent of the other variables considered. 33 The model is a little more complex as it includes two lags. However, the modelling login l| the same regardless of the number of lags specified. A Wald (Footnote continued on page I Hi [132] i lection Survey Honoured ■1 Granger causal model ol (he interrelationships lielween political mood and trust in key political institutions, 1996-2006 (quarterly) NnlKlnclion with (hp political Trust in the President Trust in the Government ' WM omnibus surveys, 1996-2006 niiiwr. infer to causal relationships that are significant (p<05). See table below for details. Vector " i"' ..ion analysis undertaken using quarterly data (q1 to q4), and refers to CVVM surveys under-tl Iwlwoen 1996 q1 and 2006 q2. I l»aln for causal independence between satisfaction with the political situation, trust in the Presi-I, ihtvnrnment and Chamber of Deputies, 1996 q1 - 2006 q2 Dependent variables ^vmlimt variables ......with the political iMiliiii (d.f. 2) m the President (d.f. 2) • ilu' Government (d.f. 2) It In the tower Chamber (d.f. 2) •j.iilnlilos (d.f. 6) Satisfaction with the Trust in the Trust in the Trust in the political situation President Government Lower Chamber 1.98 7.21** 1.96 3.17 3.00 .47 3.06 2.93 9.89** 1.99 7.02** 2.56 6.89 17.83** 12.12* 27.91*** tO «" |) .05 *** p<.001 (two tailed test) mi testing for Granger causality the null hypothesis is that the coefficients (plural because of lags) ■imcific independent variable are jointly equal to zero. Consequently, a series of restricted and un-I models are tested. The Wald test assesses whether the unrestricted estimate of a coefficient plflcantly different from a restricted estimate using a chi-square distribution with the degrees of |m equal the number of model restrictions tested. Here the degrees of freedom correspond to the ■I nl lags for which the Wald tests are calculated. For example, when explaining trust in govern-|||' lulling past values of this variable, i.e. with a lag of two quarters or six months, this improves in ."inificantly [chi-square (28,2) = 7.21, p=.03]. [133] rheory, Data and Analysis i mm.itoti buiviiy Hes0nrch other three variables. The VAR model is estimated nsiii)' < >rdinary I -east Squalen (OLS) regression where the interrelated links between the variables are modelled with no a priori expectations.34 Equally important, a VAR model facilitates using the statistical concept nl Granger causality to investigate the relationships between the political m.000 and trust measures (Freeman 1983). In simple terms, Granger causality is in ferred from the fact that past values of both the dependent and independent variables determine the current value of the dependent variable. The past can shape the future, but not vice versa. This temporal constraint facilitates makinf statistical (Wald) tests of causality, in terms of direction, reciprocity and ind@ pendence. The results of the VAR model presented in Figure 3.3 reveal that the political mood measure, i.e. satisfaction with the political situation, is Granger causally independent of all the trust indicators. Moreover, political mood only has a gig nificant effect on trust in the Government. Thereafter, trust in government detet mines trust in parliament; and this in turn shapes trust in the President. The pat tern evident at the top of Figure 3.4 reveals (1) no reciprocal causation, and (2) h hierarchical relationship between the trust questions examined. Uni-directional causation and independence suggest that the set of four po litical measures do capture different facets of Czech citizens' perceptions of M tional politics where the different CVVM questions should not be considered ni manifest indicators of an underlying latent political mood measure: one plausi ble interpretation of the pattern evident in the centre of Figure 3.3. The direr tions of the causal arrows at the top of this figure suggest that the political mood question (satisfaction with the political situation) is independent of attitudes ol trust in political institutions. However, changes in political mood do shape cm zens' sense of trust in a very specific hierarchical way. Changes in mood inllti ence trust in government which in turn shapes trust in parliament that in turn hui an impact on trust in the President. These results imply that the pattern evident in Figure 3.3 has a very speed ic structure where the general public mood is channelled through attitudes ol (Footnote continued...) test of lag restrictions indicates that some variables (trust in the Pi..... dent and Government) require a lag(2) specification. More details of the model estimation .mil diagnostics are given in Appendix 3.2. 34 Using an OLS estimator with non-stationary data is problematic because of the dangoi m making invalid inferences. Time series variables should be stationary (i.e. mean, variance; .mil covariance of each variable should not depend on time indicating the presence of an undorly ing (non)linear trend). A standard strategy to ensure stationary is to first difference the data, i | to estimate change per unit time. Unfortunately, differencing destroys information such as lony run relationships (Beck 1991: 67-69). However, with VAR use of non-stationary variables when the goal is to identify relationships rather than accurately estimate coefficients is a valid mm cise (Freeman, Williams and Lin 1989). [134] ......i die (to vein i neu I, I,owe i < 'ham In i and I'icsidenI in sequence. One ol the n mi implications to be taken 11oni this lime sei ies (VAR) analysis here is thai ''•I.....,"li there may be strong correlations between the public's political mood i"isl variables; they refer to different political attitudes within the Czech '•' loiale and should be seen as conceptually different measures. In this sub i. lion, the focus has been on the large number of inter-electoral surveys and 'i ippmg the evolution of political attitudes. Politics of course is primarily driv m I- actions; and for this reason it is very important to consider in the penul- ......1 section of this chapter data reflecting Czech's actual electoral behav mi li is therefore appropriate at this juncture to turn our attention to election i 'I Aggregate electoral data analysis research l\ 1111111 political science there is a long tradition of using official or aggregated I lli'i'lion results as these have been available since the progressive extension ol ilit lianchise in Europe and elsewhere since the late eighteenth century. These I tint it are important because they are an accurate record of citizens' political be-I hitvioiir; and it is possible to use them to make spatial (inter-constituency) and <■ mporal (inter-election) comparisons, and thereby explore the patterning and I dynamics of electoral behaviour within states. The construction of pan-Europe- 111 historical databases of constituency level election results have promoted this I Itream of research and key themes such as voter turnout, partisan support and the I (mergence of national political systems (Caramani 2000, 2004). 11 ere our focus is the organisation and use of Czech electoral statistics. It is 111(portant at the outset to provide some practical information regarding how offi-I t'lnl electoral data are archived and organised. Within the Czech Republic the or-I gimisation of elections is the responsibility of the Interior Ministry. The official I fVNiilts of all elections since 1990 are available from the Czech Statistical Office I (f'Sl I). Its website (http://www.volby.cz/) has data for all national elections since i ''ML At this website, the user may explore voter participation and party choice ii i he following levels in ascending order of size. I 1. Okrsky or precincts (n« 15,000) 2. Obce or communities (n~6,000) 3. Soudni okresy or judicial districts, sometimes also referred to as coun- ties (n=76) [135] 4. Kraje or region, typically also a constituent ) in lower chamber elee lions (n-14) 5. Národ or country (n=l) This geographical system of administration has exisled in this general form Ibl close to a century and a half. In 1869, the regions of the Austro-Hungarian líni pire forming part of the Austrian part of the Dual Monarchy (formed in 1861) were organised into a county system that was used for local administration ami electoral purposes (Mills Kelly 2007). The nature and composition of this spalial hierarchy have modified over time because of demographic and political chain' es. Nonetheless, there is a reasonable level of continuity within this schema I o explore social, political and economic change within the territory of the conleni porary Czech Republic to allow some analysis of electoral stability and change over time. However, it is important to be aware that there is some debate regard ing the classification of communities (obcí) in the study of local party systems (Hoskovec and Balík 2010). 3.9.1 Historical electoral data and analysis Information and data on national elections (the Lower Chamber and Senate) dm ing the First Republic (1918-1935) and immediately after the Second World War (1946) are given in a two volume study by the Czech Statistical Office (Kuklík et al. 2008).35 Between 1920 and 1946 there were four lower and senate chain ber elections held simultaneously in 1920, 1925, 1929 and 1935. A large number of parties (16 to 22 per election) competed for 300 seats in the Lower Chamber and 150 seats in the Senate; where throughout the fifteen year period more than 50 parties competed for seats. The large number of parties reflected the ethnic nature of the Czechoslovak state where there were in essence four party sys tems generally reflecting left-right policy orientations among Czechs, Germans, Slovaks and Hungarians. Minorities such as the Ruthenians in the sub-Carpathi an and the Poles in the Slezsko regions were never large or organised enough to constitute pivotal segments of the electorate in coalition bargaining. The fissiparous effects of using a party list proportional electoral system in an ethnically divided state were attenuated by informal mechanisms such as (a) con sensus agreements among the five main party leaders known as 'pětka', and (b) 35 Elections for the office of President were undertaken within the two chambers of parlia. J ment. The same selectorate has been used since 1990 in electing the head of state. For more de tails on the history of Czech presidential elections between 1918 and 2008 see, Tabery (2008) Slovakia changed its rules following the dissolution of the Czechoslovak Federation and it has a popularly elected president. Popular elections for the Czech presidency will take place for the first time in early 2013. Din.......ii . hiokcrccl through President Masaryk's oil ice in an inlorinal system ii ' 111 .id or I he cast I e (I .uchherl 1991: .'91; ()r/.olT 2009)."' 1 lowever, use i ili. . iwo mechanisms to juslify more efficient govemmenl decision making tiiiiiii' unit-s of crisis is seen to have undermined popular support in party politi, I In it is some evidence of this feeling in the decision to limit the number of UMI'ticN (6 and later 8) allowed to compete in the general election of 1946 under iln I i.i 11 it work of the National Front (see, Kaplan 1997). In hl.liiion, there were municipal elections where the first was held in 1919 l<..... in the I irst national elections, which is a little unusual. During the First Re- I'ulili. 111.1 under communism electoral participation was mandatory; and con-.. 111. i ■ 11 ■, voter turnout rates were typically very high (> 90%).37 Bicameral sys- i. in .in ii:. ually justified on the basis that each chamber has a different electorate iih 11 ml i as ling priorities and interests, and will thus generate different election ni> "int . I )uring the First Republic, the qualifications for voting and being a ii 1.11. l,i u- in the upper and lower chambers were different; however, the election ni......es as Table 3.4 demonstrates were often close to being identical. Mitse similar election outcomes, as noted earlier, may have reflected the ......• filmic and left-right cleavages in Czechoslovak society, but they also en- iiii J lhal the Senate never adopted a sufficiently independent position to endear ii. .-II to lhe Czechoslovak electorate. Notwithstanding these intrinsically impor-i mi I eat tires of electoral behaviour during the First Republic such as the relative importance of ethnicity and class on vote choice, as explored by Kopstein .i. I Wiitenberg (2009); one of the main reasons for studying historical electoral i ii i in the Czech Republic is to test the hypothesis that voting behaviour exhib-lls considerable stability.38 \n examination of the stability of voting patterns for four 'traditional parties': iln People's Party (ČSL, later KDU-CSL), the Socialist Party (ČSNS, CSS), the Social Democrats (ČSSD) and the Communists (KSČ), in the first post-commu-iii i elections in June f 990 reveals considerable similarity with the past. The pat-h i ns of support evident in Figure 3.5 for the Czechoslovak People's Party (ČSL) suggest a strong regional basis of partisan support. Often this party's support for policies that match with Catholic social democracy led scholars to conclude the .palial patterning evident in Figure 3.5 reflected the Roman Catholic orienta- iii Pětka and Hrad will be discussed later in the introduction to chapter 5. i/ the minimum age for voting age was initially 21 years, but this was later reduced to 18 yi'iirs. There were also restrictions on the minimum age for candidates for various types of elec-11.>ns. For details, see Broklová (1992). ill Some have argued that party competition in the First Republic did not take place in a single Czechoslovak party system (Kyloušek 2005). There were in fact distinct Czech, German and Slo- .ik party spaces and voting patterns where ethnicity and left-right orientation determined party choice. An analogous pattern is evident in contemporary Belgium. [136] [137] iiinoiy, li.ii.i nul Mn.iiy-.i-. Tablo 3.4: Conip.iii-.on ol parly support in the lower ,inil upper chambers (luring th<> hirst Hepiihhi: (191H 1938), per cent Year / Chamber IH.'O 1921 7929 lil.lt: Party Lower Upper Lower Upper Lower Upper Lowci Uppm CSDSD 25.7 28.1 8.9 8.8 13.0 13.0 12.5 l.'i. CSL 11.3 11.9 9.7 10.1 8.4 8.7 7.5 1 1 DSDAP 11.1 11.4 5.8 6.0 6.9 6.9 3.6 I / RSZML 9.7 10.1 13.7 13.8 15.0 15.2 14.3 lr with vote switching data across a pair of elections (Chamber Elections '002 and European Elections 2004) identified four political regions as shown in ľ"\ \.2. It should be noted that ecological inference refers to statistical methods u cd to estimate likely individual level behaviour from aggregate level data. I In se methods, as will be discussed later in Section 4.1 of Chapter 8, depend i ideally on being able to make assumptions about how individual level votes are •I.' Some of this data and related books are available at: http://www.czso.cz/csu/edicniplan.nsf/ .1 in.ll/ep-4#42 (accessed 22/02/2012). [141] Ihoory, Ualn and Analysis Box 3.2: Evidence for local political cultures wiiiiin (in- Czech Republic A common theme in Czech electoral history is the importance of region, i»......g I lie existence ol ('/cell oslovakia it was common to refer to the distinctiveness of areas of ethnic majorities, e.g. Czech, (iermaii, Slovak, Hungarian, Polish, etc. Within the Czech Republic there is frequent reference to cultural ilillei ences between Bohemia and Moravia. Consequently, it is net surprising to find that research using nlll cial election statistics often emphasises the spatial distribution of party support as discussed in this chat) ter and shown in Figure 3.5. This empirical evidence suggests that there are distinct regional patterns In the Czech Republic and hence the basis for local political cultures. In contrast, comparative analyses ill parly system nationalisation reveal that the Czech Republic has relatively low levels of regional vol inn implying that local political cultures are not that important. Source: Election Statistics, Czech Statistical Office (http://www.volby.cz/); Lyons and Linek (2010: W11 Note that the classification of counties and county towns is based on a hierarchical cluster analysis nt the Lower Chamber election results of 2002 and European elections of 2004. The regions are labelled m follows (I) Bohemia and urban Moravia (dark grey); (2) Rural Moravia (white); (3) Prague (black); (•!) Northwest Bohemian borderland (light grey). Districts with different coloured solid circles at their centre indicate areas where there were urban/rural differences. Lyons and Linek (2010) examined this puzzle by employing an alternative approach to the statistical aunt ysis of aggregate level election data. An ecological inference technique was used to make estimates of vote switching behaviour at the individual level across a pair of elections. One important step in this pro cess is the identification of regions where voting patterns are similar. The results of this analysis presented in the map above reveal the existence of four distinct regions or political cultures in the Czech Republic One interesting feature of this analysis is that the broad division of the country into Bohemia anil Moravia simplifies a more complicated situation where urban/rural divisions are also important. In adill tion, the impact of history is evident in the fourth region on the map. The Northwest Bohemian bordei land covers much of the territory associated with the German speaking Sudetenland. This area was reset tied after the Second World War following the forced removal of the local German population. The new settlers' community structures were not only different to the German communities; but have remained distinct when compared to the rest of the country. This is especially evident in the persistently low lev els of electoral turnout. The inductive approach to the identification and study of local political cultures using aggregated elee tion statistics represents an interesting and important stream of research. Future work employing longer time periods and data from a broader range of election types will undoubtedly add greater detail to the map shown above; and may perhaps also provide insight into the dynamics of change in Czech politl cat culture. Election Survey Rosofirr.il - lied to form the patterns observed (Aehen and SI lively 19X9; Wakefield ""I liccdman el al. 2008: 83 104). i his ecological inference work is theoretically interesting because it shows 'ii.....a high nationalised party system such as the Czech Republic, where parlies obtain approximately the same level of support in all constituencies as shown III Table 3.5. The presence of non-uniform electoral swings shows that voters do NiM view elections in the same manner as some advocates of the party system na-.....ulisalion thesis contend (Caramani 2004: 39^40). i .ii' i '>: Party system nationalisation in the Czech Republic llmilinn type and year General Election (GE) EP GE EP 1990 1992 1996 1998 2002 2004 2006 2009 or .91 - - - - - - - 0DS - .92 .91 .89 .90 .93 .91 .92 Cbsd .76 .88 .91 .93 .95 .95 .92 .91 KfiCM .95 .95 .94 .92 .90 .91 .91 .91 KOU csl .81 .70 .77 .80 .87 .70 .76 .70 b/ .81 - - - .92 .90 .88 .79 11 ID-SMS .38 .40 - - - - - - 'ii rsc/rms - .84 .86 .87 - - - - OfJA - .73 .86 - - - - - " . US DEU - - .89 - - - - ViiIim turnout 1.00 .98 .98 .98 .97 .94 .97 .94 A lean total score .80 .80 .89 .90 .92 .89 .89 .86 §twce: Linek and Lyons (2010: 380); Election Statistics, Czech Statistical Office (http://www.volby.cz/) Nulu lluit the estimates are 'inverted' Gini coefficients of party support weighted according to the size of I.....nit of analysis for all Lower Chamber Elections (or General Elections, GE) and European Parliament ■Inctions (EP) since 1990 within the Czech Republic. The units of analysis are electoral constituencies 11 WOO -1998, N=8; 2002-2009, N=14). These units are not constituencies for EP elections as the whole (fiuntry is a single constituency. When smaller units are used instead of constituencies (76 counties + 15 hfiiiue units, N=91; or counties divided into urban and rural areas + 15 Prague units, N=159), the results Din on average lower by .02. KDU-ČSL and US-DEU ran in 2002 as electoral coalition under the name Hualice (these figures are in the KDU-ČSL row). The mean total score is the arithmetic mean for all par linn Dud voter turnout and provides an overall measure of party system nationalisation. /■■<»/. OF: Civic Forum (umbrella movement); ODS: Civic Democrats (rightist); ČSSD: Social Demo-BINIh (leftist); KSČM: Communist Party (extreme left); KDU-ČSL: Christian Democrats (centre-right); SZ: .">n Party (centre-right); HSD-SMS (a small regional party in Moravia) and SPR-RSC/RMS: Republican ......(nationalist); ODA/US-DEU: Union of Freedom (rightist). [142] [143] I lection Survey Research An alternative use of (Bayesian hierarchical) ccologu al mlereiice will) ('/cell electoral and census data for the 1929 and 1935 lower chamber elections has ex amined if increased inter-ethnic contact in local communities (obec) was asso ciated with greater support for liberal parties. Kopstcin and Wittenberg (2009) find that the link between the ethnic composition of communities and non-elhnu voting was weak as other intervening factors also played an important role. i ,aU i work by Gregor (2012) using the same ecological inference technique examined key implications of Gregory M. Luebbert's (1991) theory regarding the Iran m tion away from democracy during the inter-war period in Europe. This study shows that this theory helps explain using voter transition estimates why tin Czechoslovak First Republic remained democratic when neighbouring couUlrica did not. The goal of this brief overview of aggregate electoral data analysis in the Czech Republic has been to highlight two central points. First, there is a wealth of data available for the analysis of electoral participation and party choice; and this resource is expanding as Czechs participate in an increasing number of typed of elections. Second, there is already a well-developed literature on aggregali electoral data using a wide variety of techniques ranging from maps to regression models and ecological inference analyses of vote switching behaviour, 'fluid there are important opportunities for integrating electoral data with map based databases using Geographic Information Systems (GIS); and use of mullilevi I modelling techniques when combining aggregate election results with indivld ual level survey data. In short, there is still much to be learned from aggregated, electoral data. Conclusion The central goal of this chapter has been to provide an introduction to the dill, i ent types of data associated with citizen elections in the Czech Republic sun. 1990. Consequently, the approach has been descriptive where the aim has been to identify and map out the most important sources of survey data based foi tin most part on representative national samples. All of these data are archived at CSDA, GESIS or UKDA and are freely available for academic use. In the Czech Republic there are broadly speaking seven types of election sin veys that focus on voters (plus candidate and party member surveys) attiludin and behaviour. These citizen election studies differ on the basis of type of sm vey and when the interviewing has taken place during the election cycle. As inn i election periods constitute most of the time observed, there are most survey dull [144] l"i ľ ' ailed party choice and vole intentions in the next election. Il should be not I Hi ii such inter-eleeloral estimates of electoral behaviour by CVVM, STEM, i ľ mm Invenio, SC&C, etc. are likely to have relatively high levels of measurc-tlienl error because most voters outside of election campaigns have limited inlbr-fo lllon about, or indeed interest in, elections. i" ihis respect, one would expect that the stability and reliability of voting i" I e rences recorded in inter-election surveys will vary systematically through ''" ' lection cycle. More specifically, the correlation between vote intentions and Allure (ami past) party choices will be greatest immediately before and after flections; and will be least at the mid-point between successive elections (Gel- .....1 -""I King 1993; Arceneaux 2006; Lyons 2008a: 82-87). Moreover, there ......reasons to think that other standard questions asked frequently in in- ' i ' lection polls such as trust in political institutions are likely to exhibit sys-i. 11i.ilie patterns that reflect such factors as (a) the partisanship of the respond-$11 vis-ä-vis the incumbent government, (b) the presence of scandals, (c) state i ilie economy, (d) methodological effects such as changed question ordering If re\ isions in the question or response format, and (e) idiosyncratic effects be-vnnd sampling and measurement error that are difficult to identify in the absence ili"nelical or a priori expectations. In sum, the variation present in inter-elec-llnn surveys must be examined carefully as some of the observed variance has its •i m proximate real world events; and the rest is due to systematic variations m public interest in politics. I lie final section of this chapter showed that the study of Czech citizen politics ! 1» not restricted to surveys. Aggregated electoral data have the distinct advantage i I icing an unbiased and accurate record of what citizens did on election day. Of 'ne, these official election results are aggregated to ensure secrecy of the bal-i |ltl: and so it is not possible to test individual level voting models. Ecological in-li n nee estimators may be used to overcome this problem, however, here much ds on the validity of the models' assumptions. In this chapter, the focus has been on the Czech Republic and electoral behav-§111 I fortunately, it is possible to adopt a much broader comparative perspective 11......r.h the use of an ever growing set of international surveys dealing with po- I'n al altitudes and behaviour. It is to this topic that we now turn to in chapter 4. [145]