A N A LY S I S DEMOCRACY AND HUMAN RIGHTS ELECTORAL COMPASS 2020 Analysis of the political landscape in Serbia Dr. Slobodan Cvejić, Dr. Dušan Spasojević, Dr. Dragan Stanojević, Dr. Bojan Todosijević November 2020 The economic left-right dimension is poorly associated with voting intentions. Still, it is evident that respondents who intend to vote for parties ideologically and politically associated with the ruling SNS tend to be slightly more right-wing, while the opposition voters tend to be slightly more left-wing. The conservative-progressive dimension is much more associated with voting preferences. The pattern of associations, however, mirrors those observed for the left-right dimension. The SNS block plus some other ideologically more conservative parties(DSS, Dveri, SRS) are relatively more on the conservative side. Voting intention for the rest of opposition parties is associated with more liberal stances. Most of the Serbian(intended) voters pass the‘voter rationality test’. They are more likely to vote for parties that are closer to them ideologically, or policy-wise, than for more distant parties.

DEMOCRACY AND HUMAN RIGHTS ELECTORAL COMPASS 2020 Analysis of the political landscape in Serbia U saradnji sa:

Contents EXECUTIVE SUMMARY 2 Political Landscape in Serbia 2 Ideological Orientation: Conservative-Liberal vs. Left-Right 2 Voters’ Preferences 3 INTRODUCTION 5 POLITICAL LANDSCAPE IN SERBIA 7 Conclusions 9 IDEOLOGICAL ORIENTATION – CONSERVATIVE-LIBERAL VS. LEFT-RIGHT 10 KK Policy Dimensions 10 KK Dimensions and the Serbian Voters 10 Policy Dimensions and Socio-Demographic Characteristics of the Voters 11 Social-Cultural Attitudes – Y Scale(Conservative-Progressive) 13 Economic Attitudes – X Scale(Left-Right) 14 VOTERS’ PREFERENCES 16 Ideological/Policy Dimensions and Vote Intentions(Propensity) 16 Ideological Distance and Vote Propensity: A Meaningful Vote Choice 18 Differentiation of Policy Profiles 20 Voting Intentions(Propensity) and Specific Policy Preferences 21 Conclusions 24 REFERENCES 26 1

ELECTORAL COMPASS 2020 Executive Summary This analytical report presents the results of research of polit-• The second line divides parties between government ical orientation of political parties and voters in Serbia in the and pro-government parties on one side and the oppoperiod preceding parliamentary elections held in the spring sition parties on another--judiciary independence, freeof 2020. The analysis is based on data gathered through im- dom of media, transparency, dual education, and complementation of complex methodology set created by Neth- petences of communal police are the key issues for erlands based company Kieskompas. Kieskompas.nl is a Vot- differentiation. ing Advice Application(VAA) that was first developed for the 2006 Dutch parliamentary election using a novel approach• Most Serbian parties belong to the left side of the politto calibrate political parties on salient issues and issue dimen- ical landscape, with only three parties belonging to the sions by analyzing their official policy documentation. Elec- right side of the spectrum(SNS, JS and PSG), while an toral Compass, as it is named in Serbia, is not an electoral additional two are one in the middle--SPS and LSV. forecasting tool. It is rather a political advisory instrument that at the same time collects data about political parties’• According to party standpoints on 34 issues, we differenand voters’ political orientation and preferences. The meth- tiated between three party blocks and 2 individual parties: odology produces a double axes political diagram in which Progressive left(post DS camp), Old Conservatives’ camp, political parties are positioned--the vertical‘social-cultural’ New Conservatives’ camp(DjB/SPAS duo), SNS and SPS. axis distinguishing conservative and progressive orientations at its poles and the horizontal‘economic’ axis distinguishing• The progressive left is the only pro-EU block, although between leftist and rightist orientation. After completing a SNS, SPS and New Conservatives both have positive and questionnaire, which was done by 36,099 people in Serbia. negative standpoints on EU. Voters can see which party they are closest to and thus discover their political preferences as well as their deeper ideo-• It was not always easy to find party positions on specific logical orientation. Based on this, researchers who manage issues--out of 34 statements that were used in KK, in the application can later analyze the way in which political only 50% cases we were able to position all(or almost parties make groupings in the political space on one hand all) parties. For every 10 statements we were only able to and where voters are located in the same political space on find standpoints for less than 10 parties. Also, the num the other. This way the proximity between each voter and ber of parties with“neither agree, nor disagree” position each party can also be calculated. was highest for sensitive foreign relations statements. POLITICAL LANDSCAPE IN SERBIA • The two KK policy/ideological dimensions are able to differentiate positions of Serbian parties, although to a modest degree. Parties’ views on the economic leftright dimension are not well differentiated. Views on the socio-cultural, conservatism-liberalism dimension are more differentiated. • The main line of division between Serbian parties deal with socio-cultural values, identities and questions relating to foreign and regional issues(cooperation with Russia, NATO membership, territorial integrity of Bosnia and Herzegovina and religious freedoms in Montenegro). Societal values(LGBT rights and traditional values, migrant issues, church influence in education) and issues of decentralization. IDEOLOGICAL ORIENTATION: CONSERVATIVE-LIBERAL VS. LEFT-RIGHT • The two KK policy/ideological dimensions are able to differentiate positions of the Serbian public, although to a modest degree. • Respondents’ views on the economic left-right dimension are not well differentiated. Most respondents are around the middle or neutral position. • Views on the socio-cultural, conservatism-liberalism dimension are more differentiated. • Respondents’ positions on the two ideological dimensions are weakly associated with gender(women are slightly more liberal and leftist). 2

ELECTORAL COMPASS 2020 • Associations with age and education are also low or insignificant. Somewhat higher associations are observed for the conservative-liberal dimension: More liberal views are associated with relatively higher education and, unexpectedly, older age. • Respondents from the Belgrade region and to a more moderate degree from Vojvodina, are somewhat more liberal compared to respondents from the other two regions(Western, Southern and Eastern Serbia). • • Support for democracy in general is more pronounced among liberals and those left oriented and less among conservatives and right-wing respondents. However, conservatives and right-wing respondents are more satisfied with current state of democracy within the coun try. • Trust in political organizations and institutions is linearly related to both the cultural conservative-liberal and eco-• nomic left-right scale. Those who are left and liberal oriented tend to express lower trust in political parties, courts, prosecutors and the military, while those with conservative and right-wing attitudes show higher respect toward these institutions. • • The data indicates that the respondents are on average relatively conservative in their views. In this sense, they are closer to the SNS, SPS and SPAS, the parties which won the elections. It seems that these parties have built a platform for communication with voters on these ideas and it has proved successful. Ideals related to the rule of law, the independence of institutions, free media and the path to Euro integration are topics that resonate• with a relative minority of respondents who probably boycotted the election. • The data also shows that respondents are on average left-leaning on the economic scale. Emphasizing signifi cance of social and economic equality, support for the vulnerable, the role of the state in income distribution. However, the parties that express similar views did not pass the census, while those most distant from these views have won the elections. It seems that the• above-mentioned issues are of secondary importance and that the ideas and policy solutions represented by the parties that won the elections, although contrary to what the majority accepts did not affect the election results. VOTERS’ PREFERENCES • The two Kieskompas policy/ideological dimensions are able to differentiate the positions of the Serbian voters, although to a modest degree. The overall impression is that the quality of democracy in Serbia would benefit from more political education and informative discussion about the policy positions of different parties. • The economic left-right dimension is particularly poorly associated with voting intentions. The exception is a 3 moderate association of likelihood to vote for SNS and right-wing leaning. All the remaining coefficients are considerably lower or insignificant. However, there are some clear tendencies. Respondents who intend to vote for parties ideologically and politically associated with the ruling SNS tend to be slightly more on the rightwing, while the opposition voters tend to be slightly more to the left-wing compared to the average respondent. The conservative-progressive dimension is much more associated with voting preferences in Serbia. The pattern of associations, however, mirrors those observed for the left-right dimension. SNS and the associated parties, including SRS, plus some other ideologically more conservative parties(DSS, Dveri) are relatively more on the conservative side. Voter intention for the opposition parties is associated with more liberal stances. Potential voters of DjB represent an exception in this picture. Their views are not associated with either of the two dimensions. It is possible that their views are not well differentiated but also that their distinctive ideological outlook is not captured by these two dimensions. The analyses showed that most of the Serbian(intended) voters pass the‘voter rationality test’. Namely, it proved that the ideological distance to a party is associated with voter propensity. In other words, Serbian voters are more likely to vote for parties that are closer to them ideologically, or policy-wise than for more distant parties. Although the associations are not of impressive strength (for DS the correlation coefficient is r=-.35, and for SNS r=-.45; these are the highest coefficients observed), they indicate that democratic representation is not absent from Serbian electoral politics. Voters tend to support parties with more similar policy profiles. This does not apply equally to voters of all parties. Some of the associations are rather low or insignificant--for instance, in the case of JS and SPAS voters. A somewhat more demanding test is to not vote for parties that differ from one’s policy preferences. This requires voters to be able to differentiate between parties who have similar positions from those occupying different policy positions for themselves. It proved that, indeed, voting propensity declines as parties have more different policy positions from potential voters. For instance, respondents are more likely to vote for SNS or SPS as their policy preferences are more different from those of DS. Likewise, voters are more likely to vote for DS and PSG the further their policy preferences are from those of SNS. So, yes, elements of rational, policy-based voting intentions are recorded by the Electoral Compass. Although the degree is not too impressive, nor does it apply to all parties but it demonstrated the basic democratic competence of Serbian voters. However, there is a lot of room for improvement.

ELECTORAL COMPASS 2020 • Positions on specific issues(i.e. not on the general di mensions that summarize the individual issues) are also associated with voting propensities. SNS proved to have the clearest policy profile, as there is the highest number of significant and relatively high coefficients for their in-• We would like to emphasize two implications of the pretended voters. They probably received a clearer message sented findings. One is the need for political education from the positions of this party. From one angle, this is and more substantial information to the public about expected, as the media is strongly dominated by the party policy positions. This is a requirement for developcontent related to this party. The opposition parties have ing a more functional democracy in Serbia. A more much more difficulties accessing to the media and open and politically balanced media is important here. therefore to promote their policy positions. But, other efforts aimed at drawing public attention towards policy positions of different parties, such as Elec• It should be emphasized that the strongest associations, toral Compass, can have an influential role. In an envi both for the governing parties and the opposition, relate ronment where there is more demand for information to the‘regime cleavage’ theme. These would be items about party policy positions, parties themselves should that express positive or negative views of the current provide more clear and unambiguous information about regime(controlled by the SNS). The highest coefficients their actual positions. refer to whether the government influences the work of the judiciary, and whether media freedom was threat-• The second important implication is that the political ened during the COVID-19 state of emergency. The oth- spectrum needs to be better differentiated. While the er relatively strong associations either express evaluation Serbian party system has many parties, in addition to of specific policies implemented by the government, or the one currently dominant party, it is clear that they are concern‘symbolic’ issues, such as the attitude towards just separated into two blocs--parties associated with Russia and the EU. The classic ideological issue concern- the government(and SNS) and those opposed to them. ing economic, redistributive issues is poorly represented, There is very little differentiation within both camps. At while ecological issues are completely unrelated to party least according to the issues examined by the Kieskompreferences. pas. Even the relatively new parties, such as DJB, NOVA, or SPAS, failed to demonstrate clear and distinctive poli• As was the case with the general dimensions, the coef- cy profiles to their potential voters. It is certainly possible ficients associated with preferences for the opposition that the study omitted issues relevant to these parties. parties are also relatively weaker(and insignificant more However, the study did include the main issues that often). But, the pattern of the relationships pretty con- were discussed during the election campaign. Again, sistently reflects the associations with the governing this implies the need for both: abilities of parties to presparties. One distinction seems to be that in the case of ent their policy positions and the need to encourage the these parties’ voters, the symbolic issues have a some- public to take those positions into account when making what stronger relative importance--especially the EU is- their vote choice. sues, the Kosovo issue and same-sex marriages. So, it seems that potential voters of the liberal opposition are differentiated by the‘Eurocentric’ and libertarian outlook, in addition to the dislike of the governing parties. 4

ELECTORAL COMPASS 2020 INTRODUCTION This analytical report presents research results of the political orientation of political parties and voters in Serbia in the period preceding parliamentary elections held in spring of 2020. The project was supported by Friedrich Ebert Stiftung Serbia and Heinrich Böll Stiftung Serbia. The analysis is based on data gathered through implementation of complex methodology set by Netherlands based company Kieskompas. The process of data gathering in Serbia was organized in cooperation with Blic newspaper and web portal(a part of Ringier Axel Springer Media AG) under the name Izborni kompas(Electoral Compass). Kieskompas.nl is a Voting Advice Application(VAA) that was first developed for the 2006 Dutch parliamentary election us ing a novel approach to calibrate political parties on salient issues and dimensions by analyzing their official policy documen tation. Electoral Compass is not an electoral forecasting tool. It is rather a political advisory instrument that at the same time collects data about political parties’ and voters’ political orientation and preferences. The methodology produces a double axes political diagram in which political parties are positioned--the vertical‘social-cultural’ axis distinguishing conservative and progressive orientations at its poles and horizontal ‘economic’ axis distinguishing between leftist and rightist orientation. After completing a questionnaire, voters can see which party they are closest to and thus face not only their political preferences but deeper ideological orientation too. Based on this, researchers who manage the application can later on analyze the way in which political parties make groupings in the political space on one hand and where the voters are located in the same political space on the other. This way the proximity between each voter and each party can be calculated, too. And that is exactly what we present in this report. After this introduction, a chapter explaining the positioning of political parties in Serbia follows. The next is a chapter about voter positioning in the same political landscape divided by the two ideological axes mentioned above. In the final chapter, we present the proximity between the parties and the voters using the same methodological tool. Kieskompas VB gathered and trained a team of Serbian experts in political science, sociology and research methodology who adjusted the contents of Kieskompas’ methodology to the Serbian political context. The adjustment of methodology assumed three major steps: 1. Selection of topics and statements through which positioning of political parties and voters in political landscape of Serbia would be conducted. 2. Selection of political parties to be assessed and positioned in the political landscape. 3. Validation of adjustment results with the wider community of political analysts, journalists and politicians. Starting in February, before the parliamentary elections in Serbia officially started and before interruption of the electoral campaign due to the Covid 19 pandemic, the expert team conducted an adjustment of the Kieskompas methodology to the Serbian political context. The expert team worked through several workshops with constant support from the Kieskompas team in the Netherlands. Independent Serbian political consultants and media representatives(journalists) were occasionally involved in these workshops to obtain external validation of the proposed content. Also, the proposed topics and statements were sent to a few of the political parties in the form of a questionnaire. This way we received comments on methodology from the political parties’ representatives and at the same time validated their positioning in the political landscape based on applied methodology. Concerning the selection of topics and statements, the expert team initially started with 44. After scrupulous research and testing, the list was reduced to 28 and implemented into a beta version of the web site. After the Covid 19 pandemic the list was extended as to reflect new issues(topics) pene trating political discourse and electoral campaign. This way the final list of 34 statements was obtained. For each of the statements there is a citation for each of the selected political parties taken from their programmatic documents or from their public statements. These citations illustrate their political standing towards selected topics and contribute to their positioning in the overall political landscape. In accordance with Kieskompas methodology, the Serbian expert team didn’t ask representatives of the political parties to declare their(current) standings through an interview or a survey but relied on their official or public declarations. That is what is meant by‘the parties gave answers to each statement’, as written on the‘Izborni kompas’ web site. The selection of political parties to be presented in‘Electoral Compass’ was based on previously agreed criteria. The expert team started from 4 recognized problems in the beginning of their work, which were: 1) a large number of political parties 2) many new parties(yet to be) registered during campaign period 3) boycott vs. participation dilemma and 4) unclear composition of the future coalitions. From the methodological point of view, it was not possible to include ALL of the parties and be registered political entities in the‘Electoral Compass’ because it would be completely confusing and inconclusive. The team had to reduce the number, but still guess which parties would take part in elections and which not, the question to remained open until the very end of registering period. And we could not include coalitions be5

ELECTORAL COMPASS 2020 Table 1 The 16 political parties included in the research 1. Demokratska stranka- DS Democratic Party 2. Demokratska stranka Srbije- DSS Democratic Party of Serbia 3. Dosta je bilo- DJB Enough is enough 4. Dveri Dveri 5. Jedinstvena Srbija- JS United Serbia 6. Liga socijaldemokrata Vojvodine- LSV The League of Social Democrats of Vojvodina 7. Narodna stranka- NS People’s Party 8. NOVA stranka NOVA 9. Partija ujedinjenih penzionera Srbije- PUPS United Pensioners’ Party 10. Pokret slobodnih građana- PSG The Movement of Free Citizens 11. Socijaldemokratska stranka- SDS Social-democratic Party 12. Socijalistička partija Srbije- SPS Socialist Party of Serbia 13. Srpska napredna stranka- SNS Serbian Progressive Party 14. Srpska radikalna stranka- SRS Serbian Radical Party 15. Srpski patriotski savez- SPAS Serbian Patriotic Alliance 16. Stranka slobode i pravde- SSP The Party of Freedom and Justice cause: a) they were not formed at that moment and b) coalitions didn’t present their political attitudes towards selected topics, only their member parties did so. Due to the reasons stated above the expert team decided to include in the‘Electoral Compass’ parties which: • Have MPs in the current call of parliament as elected in previous elections, meaning the original parties and not MP clubs and/or parties that emerged meanwhile due to the split of the initial parties entering the parliament • Were estimated by the independent voting polls as having a realistic chance to win more than 3% of the electoral body(the census for this year’s election) without forming a coalition • Are not national minority parties because minority parties participate in elections by the positive discrimination rule and thus would be impossible to realistically compare to other parties in the‘Electoral Compass’ political landscape • Have developed programmatic documents or public statements that sufficiently allow for credible position ing in the political landscape in accordance with Kieskompas methodology Applying methodology described above, the expert team in cooperation with the Kieskompas team and Blic, finally launched a platform that sets 16 political parties in the political landscape based on their standings towards 34 topics/ statements. By the end of the Electoral Compass a total of 36,099 voters answered the questionnaire containing the same 34 statements and thus positioned themselves within the political landscape and were able to learn from the web site which of the presented parties was closest to them. The site did not give advice to voters on which party to vote for. Table 2 The 34 topics used for positioning of the parties and voters in political space • EU accession • NATO membership • Rusia vs. EU support • Friendship with China • Independence of Kosovo • Religious protests in Montenegro • Same sex marriages • Settlement of non-European immigrants in Serbia • Taxation of religious institutions • Letting immigrant workers in • Governmental support to patriarchal values • Government controlling judiciary • Decentralization • Direct elections • Subsidies to foreign investors • Dual secondary education • Neo liberal labor market • Free housing for young experts • Increase of pensions only if following economic growth • Public investments in health system • State subsidies to public enterprises • Increase of minimum income • To ban communal police •‘Helicopter money’(100 Euro) as a support measure during Covid 19 • Citizens’ health is priority over economic growth • Sufficiency of government’s economic measures • Support to martial law during the first wave of Covid 19 pandemic • Overly restrictive rules for elderly during pandemic • Elections should be postponed • Media freedoms restricted during pandemic • Citizens responsible for pollution • Environment protection more important than economic growth • Pandemic is a good opportunity for economy of Serbia 6

ELECTORAL COMPASS 2020 POLITICAL LANDSCAPE IN SERBIA Ideological positioning of Serbian political parties was based on 34 questions distributed onto two scales(as described). The ideological landscape revealed several expected characteristics and some new findings. Political parties are more differentiated based on cultural values(vertical scale). This has usually been explained by the predominance of identity based issues in contemporary political discourse(some of these originated from the breakup of Yugoslavia, while others can be perceived as usual transitional dilemmas), these are in line with other post socialist transitions(Elster, Offe, Preuss, 1998). Another reason lies in the lack of articulated economic policies(represented in the horizontal scale) by the political parties, or by other usual actors in these fields--e.g. trade unions or capital owners’ associa tions. More precisely, the questions that were the most salient and divisive in Serbia in the last 20 years are: dealing with cooperation with the International Criminal Tribunal for the Former Yugoslavia(ICTY), the future status of Kosovo and/or European integration 1 . At the same time, some of the most salient issues, like the Kosovo issue, still lack clear understanding from the political parties’ standpoints and any classifying parties should as for gradual and careful assessment? The meaning here is unclear. Economically left As stated above, some extremely important issues seem non divisive--foreign policy on EU integration and the future status of Kosovo are the top picks among those. In similar fashion, anticorruption and environmental statements produce low degrees of polarization between parties, as well as most economic issues. Examples of statements that produce moderate polarization are debt collection issues(who should collect debts, state or private agencies), selling the land to foreign companies and privatization of large old public enterprises that are constantly producing debts. Figure 1 Political landscape Progressive Economically right Electoral Compass provides us with insight into the level of polarization created by these proposed statements(Spasojević, Stojiljković, 2020). Highly divisive topics are related to foreign and regional issues(cooperation with Russia, NATO membership, territorial integrity of Bosnia and Herzegovina and religious freedoms in Montenegro), social values(LGBT rights and traditional values, migrant issues, church influence in education) and issues of decentralization. The largest number of statements divided parties between the government and pro-government parties on one side and the opposition parties on the other were: judiciary independence, freedom of media, transparency, dual education and competences of community police. Most of these issues are related to the quality of democracy in Serbia and these issues serve as the foundation of regime cleavage(increasingly becoming the most important line of division in Serbian politics). Conservative At first glance, it seems that party system presented by Elec toral Compass has left wing bias, since only three parties belong to the right side of the spectrum(SNS, JS and PSG), while an additional two are in the middle--SPS and LSV. However, this landscape is not that different from a general understanding of the Serbian party system.“Left wing bias” is usually explained by Serbian median? Voter(s) being slightly pro-state redistribution and/or state dependent 2 . Left wing biases are also more visible due to the absence of liberal parties or parties that are perceived primarily through liberal economic policies. 1 Spasojević, Dušan i Stojiljković, Zoran.(2020). Između uverenja i interesa. Fabrika knjiga. Beograd. 2 For example of this understanding see Stojiljković Zoran(2009), Partijski sistem Srbije, Službeni glasnik, Beograd of edited volume Kako glasam, zašto glasam, ako uopšte glasam(edited by Stojiljković, published by FES in 2012) 7

ELECTORAL COMPASS 2020 In the Electoral Compass landscape, this left wing bias is emphasized even more due to the following issues:(1) higher number of parties in the opposition camp are produced to a large extent by the atomization of parties descending from the Democratic Party and was traditionally leftist,(2) this shift to the left is stronger because most of the parties on the left side of the diagram are oppositional; oppositional parties, at least in the Serbian case, always tend to promote more state support for different group,(3) also, some of the currently salient and polarizing issues(e.g. dual education or 100 euro government support to all citizens) have shifted parties more to the opposite poles of the spectrum. Political parties can be grouped into 3 camps with two individual parties staying out of the groupings: 1. Progressive left(post DS camp), 2. Old conservatives’ camp 3. New conservatives’ camp(DjB/SPAS duo), 4. SNS and 5. SPS. The progressive left(or post-DS) camp includes the following parties--DS, NOVA, SDS, SSP,NS, LSV and PSG and gathers them around more or less strong pro-European and progressive standpoints with slight inclination towards state redistribution in economy. Most parties in this block are partial or complete spin-offs from the Democratic Party(DS), their coalition partners(LSV and NOVA) or newly founded parties (PSG) have already cooperated with the democrats(Janković’s presidential campaign in 2017). The core of the group includes very similar ideological parties--DS, NOVA, SDS and SSP. The block also includes the less EU enthusiastic party NS (led by Vuk Jeremić) that is somewhat specific in foreign pol icy issues and more progressive and more liberal than other parties like LSV etc.(positioned in the middle on the economy scale) and PSG, which attempts to occupy the position for classic liberal parties. Regardless of partially different positions on EU integration between parties of the progressive block, this issue is still the dividing line between conservative and progressive camps in Serbia--both on the party level and on the level of individual voters. Figure 2 Views on EU membership by progressive and conservative blocks‘ voters the DS was the driving force of the economic liberalization in the first decade of post-socialist transition; their voters were also more inclined toward pro-market measures and being less dependent on state budget support compared to voters of SPS and SRS/SNS for example(transitional winners vs. losers paradigm). However, these parties(DS primarily) always tried to present themselves as social-democrats. After the change of government in 2012, DS shifted more to the left 3 , although those positions were less articulated by party program and more by leaders’ statements. In the case of SDS, which was founded later than other DS descendants, this shift was more clear and party program and statements were more aligned. The‘Old’ conservative camp consists of 5 parties--SRS, JS, PUPS, Dveri and DSS, although neither of these are homogenous groups, SRS has been the strongest party promoting nationalism and far right politics since the renewal of party pluralism in Serbia; JS is a regional party(central Serbia) with similar standpoints to SRS. In contrast to SRS and JS, Dveri are more driven by social conservatism and orthodox Christian values, although they also share some nationalistic views. Dveri are also more pro-state redistribution(leftist) compared to other parties from this group; due to their protectionist claims, Dveri represents the most leftist party in the entire Electoral Compass field. The additional difference between these parties is based on standpoints on the current Serbian government--JS is a part of the government coalition, while SRS shows more understanding for government policies compared to other oppositional parties. Dveri is strongly against the government and share the views of progressive parties in this regard. DSS was one of the two largest parties in the first post Mi losevic decade; after the change of leadership(Kostunica re tired after 2008), DSS tried to position itself as a moderate conservative party--the outcome became the most progressive position among the conservative camp, moderate(compared to other parties) pro-state redistribution ideas and positioned between the government and the opposition (although more critical on government compared to SRS). Finally, PUPS is classified in this group, but they have quite a reduced program and it was not possible to find their stand point in relation to many of the statements in Electoral Compass. Therefore, their conservative character is based on very few indicators and PUPS is more identified by their economic attitudes(related to topics relevant to retired people). The new conservative camp includes two parties--Dosta je bilo and SPAS. Both parties have roots in the progressive part of the landscape.‘SPAS’s president was a member of DS and DjB was perceived as a party of similar(liberal, pro-meritocratic) ideological profile. However, both parties changed their positions and became more conservative, although driven by new ideas and issues instead of old nationalistic conservatism(represented by SRS, for example). SPAS preserved Position of core group of the progressive left(DS, NOVA, SDS, SSP) on the economic scale is somewhat debatable(left wing bias) and it should be taken with precaution because 3 Stojiljković, Pilipović i Spasojević, ur.(2014) Izborne ponude i rezultati--evropski izbori i izbori u Srbiji 2014. godine, Beograd: Konrad Adenauer Stiftung.; or journal Oko izbora, no. 19, 2014. 8

ELECTORAL COMPASS 2020 similar foreign policy standpoints as DS, but adopted more flexible standpoints on cooperation with China and different attitudes towards Montenegrin law and religious freedoms. SPAS also became a strong anti-immigrant party. In contrast to other opposition parties from the progressive block, SPAS was less critical toward the government policies during the Covid 19 outbreak(issues of freedoms and rights during the martial law, including media freedom). Interestingly, together with some other new parties, SPAS supported the idea of a preferential voting system, probably because their leader was a well known athlete. Dosta je bilo(DjB) transformed itself rapidly during the 2016-2020 parliamentary mandate into a strong anti-EU and anti-NATO party, with harsh anti-immigrant attitudes. However, each party kept their progressive understanding of democracy and media freedoms, and that is why it has been classified between the blocks of conserva tive and progressive parties. In terms of economic policy DjB has mixed ideas--some pro market(on the role of state in employment policies or subsidies for old companies that are creating debts) and some pro-state redistribution(on dual education or budget support to companies during the Covid 19 crisis), with leftist points on more state budgeted investments in the health care system and more aid for companies after the Covid 19 crisis. Therefore, their moderate pro-redistributive position is also contextual. Defining ideological positioning of SNS was quite hard. This party has a unique position due to its size(between 40 and 50% of the electorate), a colorful electoral and an interesting position on the economic scale in relation. However, SNS’s right wing position is not based in liberal economic policies (or some fundamental liberal values), but it is more contextual and related to government policies--support for the state subvention for foreign companies,‘helicopter money’ intervention after the breakout of Covid19 or dual education. When it comes to general issues, like employment policies, SNS tends to be as close to center as possible. This is also clearly visible regarding conservative vs. progressive issues, because SNS again tries to be as central as possible and to avoid any radical ideological characteristics. This goes beyond usual understanding of catch-all policies and clearly represents SNS’s long term strategy. Finally, Socialists are another party that profit from centrist positioning and significant coalition potential. Although they are a pro-EU party, their support for conservative government policies during the pandemic crisis generated their position below the center in the ideological landscape. It is also interesting to note that the Socialist Party does not promote strong leftist ideas, although some of their economic positions are a direct consequence of coalition with the much stronger right-wing party(SNS). However, SPS provides strong support for different kind of subsidies(including ones for direct foreign investments and immigrant workers, as well) and state intervention employment policies, as well as further financial support from the state budget to old non-privatized companies. CONCLUSIONS • The two KK policy/ideological dimensions are able to differentiate positions of the Serbian parties, although to a modest degree. Parties’ views on the economic leftright dimension are not well differentiated. Views on the socio-cultural, conservatism-liberalism dimension are more differentiated. • The main line of division between Serbian parties deals with socio-cultural values, identities and questions relating to foreign and regional issues(cooperation with Russia, NATO membership, territorial integrity of Bosnia and Herzegovina and religious freedoms in Montenegro), social values(LGBT rights and traditional values, migrant issues; church influence in education) and issues of de centralization. • The second line divides parties between government and pro-government parties on one side and the opposition parties on another--judiciary independence, freedom of media, transparency, dual education and competences of communal police are the key issues for differentiation. • Most Serbian parties belong to the left side of the political landscape, with only three parties belonging to the right side of the spectrum(SNS, JS and PSG), while an additional two are one in the middle--SPS and LSV. • According to party standpoints on 34 issues, we differentiated between three party blocks and 2 individual parties: Progressive left(post DS camp), Old conservatives’ camp, New conservatives’ camp(DjB/SPAS duo), SNS and SPS. • The progressive left is the only pro-EU block, although SNS, SPS and New conservatives have both positive and negative standpoints on EU. • It was not always easy to find party positions on specific issues--out of the 34 statements we used in KK, in only 50% of the cases we were able to position all(or almost all) parties. For 10 statements we were able to find standpoints for less than 10 parties. Also, the number of parties with“neither agree, nor disagree” position was highest for sensitive foreign relations statements. 9

ELECTORAL COMPASS 2020 IDEOLOGICAL ORIENTATION: CONSERVATIVE-LIBERAL VS. LEFT-RIGHT KK POLICY DIMENSIONS Kieskompas(KK) model frames political space by two dimensions. The horizontal one represents the traditional socio-economic left-right policies. The left side involves policies aimed at reducing social and economic inequalities, support for state redistribution, and similar topics. The right side of the spectrum emphasizes the market allocation of wealth and opposition to state intervention in the economy. means that Serbian respondents tend to endorse middle positions, rather than extreme in any direction. In cases of strong ideological polarization, one can observe increased numbers towards the extremes of the dimension. On these dimensions in Serbia, citizens do not have strongly polarized views but rather centrist, neutral or undifferentiated views. Figure 3 Distribution of the two ideological dimensions The vertical dimension of the KK model concerns cultural, life/ style issues, and policies. On the conservative side, one can find preferences for policies aimed at promoting traditional family models, authoritarian social relationships, opposition to unconventional lifestyles, but also nationalism, traditionalism, ethnocentrism and similar values. The social-liberal pole of the dimension involves a preference for liberal social policies, tolerance of unconventional lifestyles, ethnic tolerance, internationalism. Public opinion research shows that voters’ political preferences are more based on the traditional/authoritarian versus liberal/internationalist dimension than on the economic leftright dimension(e.g., Kuzmanović, 2010; Mihailović, 1991; Pantić, 2003; Pavlović, Todosijević& Komar, 2019; Todosije vić, 2013). Moreover, when Serbian respondents describe themselves ideologically, labels such as traditionalist, patriot and nationalist are more politically relevant compared to leftright self-identification(Pavlović& Todosijević, 2016). In this section, we will explore Serbian participants‘positioning on the KK dimensions, and how they relate to basic social background variables, some of which are typically associated with the ideological dimensions. However, there is a distinct difference between the two distributions. The Left-Right dimension is more centered, or more narrowly distributed around the center. This means that the majority of the public ends up around the middle of this dimension, rather few exhibited decisively leftist or rightist policy preferences at this time. On the other hand, the distribution of voters’ political preferences on the socio-cultural dimension is more widespread, which demonstrates more differentiated views on policies relating to the conservative-liberal dimension. The spread seems to be particularly extended in the liberal direction. KK DIMENSIONS AND THE SERBIAN VOTERS First, we look at the distribution of the two dimensions, based on the answers from about 30,000 respondents who participated in Electoral Compass in Serbia. The following graph shows the distribution of the two dimensions(kernel density graph). We can first observe that both dimensions have a clear tendency towards the center, resembling the normal distribution. On both dimensions, this Undifferentiated policy views(or ideological orientations) might seem somewhat undesirable from the perspective of normative political theory. If voters do not show consistent and differentiated policy preferences, it is difficult for parties to develop and maintain clear ideological profiles. In the end, it might result in less than optimal degree of policy representation and leave the public unsatisfied with the way democracy works. The fact that the public’s own undifferentiated policy profiles contributed to such outcomes is not necessarily a great comfort. 10

ELECTORAL COMPASS 2020 A more positive interpretation of the findings would point to the similarity of the above picture to the distribution of party positions on the two dimensions. Party positions, based on the Electoral Compass expert team’s evaluation of party programs and public statements, revealed a similar picture, with a more differentiated cultural dimension compared to the left-right dimension. This similarity could be taken as a sign of corresponding or even successful representation of public policy preferences by the political parties. One could indeed argue that this is just a representation of the lack of differentiation. In subsequent sections, we will take a closer look at the specific rela tionships between voters and party positions. The following analysis examines the relationship between the two KK policy dimensions. The KK model conceives them as two independent dimensions, although in practice they may be related. A very high correlation between them would question the need to use two instead of a single dimension. In Serbia, the data shows that the two dimensions are significantly and negatively correlated(r=-.23 among male re spondents, and r=-.28 among females; in both cases p<.001) 4 . This means that respondents with more right-wing economic views tend also to hold relatively more conservative socio-cultural views. This was an expected result given that these two scales have some common assumptions, i.e. cultural conservatism often goes hand in hand with nationalist ideology, while the economic left-wingers tend to endorse slightly more liberal socio-cultural preferences, which is a characteristic found in other countries as well(e.g., Cochrane, 2013; Oesch& Rennwald, 2018). The following graphs show the joint distribution of the two dimensions, separately for male and female sub-samples. We also asked the respondents to position themselves on the classic ideological self-positioning scale(0-10) from political left to political right and found that both KK scales presented above correlate with it. The conservative-liberal scale(X) shows relatively strong connections and inversely correlates with ideological self-positioning(r=-, 423, p<, 001), mean ing that the more respondents who are conservative of the more conservative respondents and vice versa. Correlation with the economic positioning scale(Y) is moderate and direct(r=, 239, p<, 001), meaning that self-positioning on the left side of the ideological spectrum is related to the left side of our Y scale while right self-positioning corelates to the right side of the Y scale. Interestingly, the links between ideological self-positioning and the scale that measures cultural constructs are stronger than the scale that measures the economic dimension of ideology, directing us to two probable outcomes. First, for our respondents on the right and left generally express cultural rather than economic views, and second, the modern left has significantly shifted its under standing of political priorities to identity policies and political freedoms rather than focusing on economic inequalities. The gender difference between the associations among the two policy dimensions is minor and is not noticeable on the graph. One could hypothesize that the association should be stronger in the male subsample. The reason is the usual find ing that men are more interested in politics and show greater political competence(e.g., Thomas, 2012). Hence, ideological and policy preferences could be more integrated among them. But, it seems that it is not the case in Serbia when looking at these two policy dimensions. Perhaps, genders in Serbia do not differ much in general. Or, it may be the case that those who took part in the Electoral Compass survey were similarly politically interested, while the difference might exist in the general population. POLICY DIMENSIONS AND SOCIO-DEMOGRAPHIC CHARACTERISTICS OF THE VOTERS Gender differences between the average levels of the two policy dimensions are significant, as demonstrated in Figure 4. 5 The differences are, however, rather small concerning the left-right dimension. Women appear to be somewhat more left leaning in economic matters compared to men. Differences in the conservative-liberal dimension are larger. Serbian women, at least those who participated in the Electoral Compass survey, proved to be more liberal than their male counterparts. In addition to gender differences, we will examine the association between the policy dimensions and several other basic socio-economic background variables. Table 3 shows that correlations of age and education with the left-right dimension are very low or statistically insignificant. So, there is a very weak tendency for age to be higher among those who express right-leaning policy preferences(r=.038). This is not a particularly unusual finding. A recent comparative study by Caughey et al.(2019) shows that older respondents tend to be somewhat more left-leaning in economic matters compared to the younger and mid-aged categories. The education of the respondents is positively related to the conservative-liberal scale, so that with the increase of the level of education, the acceptance of liberal views increases and the acceptance of conservative views decreases. Barberá (2017), for instance, reports such findings in Germany. This finding indicates that the educational process influences the liberalization of views in the field of culture and politics. Edu cation, on the other hand, is not related to economic views (left-right), indicating that the relationship to systemic and personal responsibilities is formed primarily on the basis of personal experiences that may be equally distributed among respondents regardless of education. 4 The coefficient is statistically significant, which is not surprising gi ven the size of the sample. But, the coefficient is not high enough to warrant the unidimensionality of the examined attitudes among the Serbian public. 5 Confidence intervals for the estimates of averages do not overlap, suggesting that gender differences are indeed statistically significant. 11

Figure 4 Joint distribution of the two policy dimensions by gender ELECTORAL COMPASS 2020 Table 3 Correlation between age, education, settlement size and economic self-positioning and two ideological/policy dimensions Economic LeftRight dimension Conservative-liberal dimension Age.038**.104** Education-.001.120** Settlement size-.020**.115** Economic self-position **p<0,001 Source: survey data -,126**-.006 Cities and urban centers are the core of more liberal and leftist ideas. With the growth of the settlements in which the respondents live, the support for liberal views increases and for conservative ones decreases. At the same time, support for leftist ideas in the economy grows and declines for right ones. This finding is expected given that larger and urban centers represent places where progressive and liberal ideas are more accepted and that smaller places and villages are surroundings in which tradition lasts longer. However, contrary to the usual findings from Western Eu rope, relatively older respondents in Serbia appear to express more liberal policy preferences. A hypothesis based on the role of political socialization might help here. Older respondents are socialized under the socialist regime, and therefore may be more inclined to still show some liberal inclinations. On the other side, younger respondents are socialized in the era of‘re-traditionalization’, including de-secularization. These two processes could explain the observed deviation of the Serbian results from what is commonly found in the literature. Further on, siding with right-wing views increases with the age of the respondents while it decreases with left-wing views. The(albeit weak) link between age and the economic right indicates that as people age, they become more prone to individual responsibilities in the labour market, withdrawal of state ownership from the economy but state interventions in favor of big capital. The country region is another important socio-demographic variable. It could be expected that in more urbanized regions, which in this case would be the Belgrade region, the public tends to be more liberal, although no very strong hypothesis can be generated for regional differences per se in left-right policy preferences. Figure 5 Average levels of policy dimensions in four statistical regions of Serbia Self-assessment of the economic standard of the respondents is related to their attitudes towards economic solutions. With the decline of the economic standard(economic self-positioning), the support for“left” solutions in the economy rises and vice versa, the improvement of the economic position of respondents is accompanied by the acceptance of right ideological beliefs to a greater extent. This finding sug gests that income-related experiences are very important in explaining adherence to the“left” that places responsibility in the hands of the state and institutions, or the“right” that insists on personal responsibility and minimal state responsibility. The results show that both dimensions exhibit different average levels in different regions but the differences are more pronounced for the liberal-conservative dimension. Respondents from the Belgrade region, and to a more moderate degree from Vojvodina are more liberal and more left-wing compared to the respondents from the other two regions (western, southern, and eastern Serbia). One reason for this 12

ELECTORAL COMPASS 2020 could be the level of urbanization but it could also be due to cultural differences. In any case, a more detailed explanation would require additional research. SOCIAL-CULTURAL ATTITUDES: Y SCALE(CONSERVATIVE-PROGRESSIVE) In this part we will present statements that reflect the great est degree of agreement with conservative and liberal ideas 6 . Conservative attitudes are those that relate to anti-immigrant and anti-NATO sentiments and heteronormative patriarchal values. The most supported statements are the following: of democracy in Serbia. Although respondents are generally more dissatisfied than satisfied with the state of democracy, the satisfaction is the lowest with those more liberal and the highest with conservatives. Figure 6 Attitudes toward democracy • Permanent settlement of migrants coming from non-European countries should be banned • Serbia should never join NATO if we want to maintain independence • Serbian government should introduce new laws that preserve patriarchal family values • Same-sex marriages should never be allowed in Serbia. The most progressive statements our respondents supported are those related to judiciary independence, media freedom, the election system and the influence of church and faithbased organizations. The most supported statements are the following: • The government essentially influences the work of the judiciary • Instead of voting only for political parties, Serbian citizens should have the opportunity to vote for individuals who will represent their constituency Trust in political organizations and institutions is linearly related to the social conservative-liberal scale . Those who support liberal views have the least trust in political parties, courts, prosecutors and the military. On the other hand, those who support conservative views have more trust in all of the above institutions. Thus, conservatives are more supportive of political parties, institutions of the system, and especially institutions of force, while liberals express far greater distrust in them. General trust in people is not related to this ideological scale, indicating to us that it does not depend on ideology but on other personal characteristics. Figure 7 Trust in organizations, institutions and people in general • Media freedom was seriously threatened during the Covid pandemic state of emergency • The church must pay taxes. In order to present the results more clearly, we transformed the classical 0-10 scale of ideological self-perception of the respondents into a three-level scale of(dis)agreement with ideological orientation. The relationship between ideological orientations of respondents and their attitudes towards democracy is interesting. Those who are liberally oriented in principle, largely support a democratic system of government. Also, with the decline of liberal orientation and the transition to conservative ideological views, support for a democratic way of political governance declines, so that those who are less conservative(below the theoretical average of the scale(3)) are more inclined to support democracy. The same pattern is confirmed in regards to the relation be tween ideological orientation and satisfaction with the state 6 Measured on five item scale(1-5, the higher the score, the higher the agreement with the statement). In the next section we will present how close voters are in Serbia to different political parties assessed in our survey. This is presented through a measure of distance between each individual voter from our sample each political party in the Electoral Compass spectrum. The value of distance ranges from 0 to 3.64. The higher the value, the bigger the distance of a respondent’s views on Y scales from the party position on the same scale. Respondents who were part of our sample were to varying degrees distant from individual political parties in their views on the social conservative-liberal scale. Respondents are on average closest to SPS, DJB, NS, SPAS, SNS, SSP, and are furthest from the SRS, LSV, PSG, Dveri, and DS views expressed on Y scale. Observing the positioning of the respondents, we see that the majority of them concentrate around the middle of the scale with a slight tendency towards 13

ELECTORAL COMPASS 2020 conservative orientations. As many as three parties(SPAS, SPS and SNS), whose programmes and statements are close to the respondents on this scale, managed to collect a significant number of votes in the elections, pass the census or achieve a convincing victory. Progressive ideas, the rule of law, the independence of institutions, the media, and the path to Euro-Atlantic integration are topics that obviously did not mobilize voters, or it mobilized them to boycott the election. Table 4 Party position on the cultural Conservative-Progressive(Y) scale and average respondents’ distance from party position Party SPS DJB NS Social Conservative-Liberal Party position on Y scale 7 -0.08 -0.08 Average distance of respondents 0.53 0.55 0.25 0.55 SPAS 0.58 0.67 SNS-0.46 0.72 SSP 0.7 0.73 SDS 0.91 0.87 DSS-0.69 0.89 JS-1.0 1.16 PUPS-1.0 1.16 NOVA 1.30 1.19 DS 1.33 1.22 Dveri-1.08 1.23 PSG 1.42 1.30 LSV 1.64 1.51 SRS-1.38 1.53 ECONOMIC ATTITUDES-X SCALE(LEFT-RIGHT) On the economic left-right scale the most important leftist issues were: • the state of the health system • the preferred level of minimum wage • the need for communal police • environmental issues On the right pole of the ideological spectrum relevant topics were: • economic protectionism •‘helicopter money’ as a measure of support during Covid 19 pandemic • pension system • subsidies to public companies. Like with Y scale, here too we have an interesting relationship between ideological orientations and attitudes towards democracy. Regardless of the ideological position most of the voters agree that a democratic system is a good form of government. On the other hand, the attitude towards democracy in Serbia is connected differently with ideology. Although respondents are generally more dissatisfied with the state of democracy than they are satisfied, satisfaction is the lowest with those on the left and the highest with those on the right. Figure 9 Attitudes toward democracy Figure 8 7 Party positions on X and Y scales Progressive Economically left Economically right Figure 10 Trust in organizations, institutions and people in general Conservative 7 Numbers indicate position of the party on Y scale as KK experts estimate. The position is expressed in values that can range from-2 to +2. Minus represent conservative side of the scale and plus liberal part of the Y scale. The higher the-/+ value, the more pronounced ideological conservative/liberal orientation. 14

ELECTORAL COMPASS 2020 As with the social conservative-liberal scale, the relationship• Respondents from the Belgrade region and to a more of trust in political organizations and institutions is linearly moderate degree from Vojvodina, are somewhat more related to the economic left-right scale. Those on the left liberal compared to respondents from the other two rehave the least trust in political parties, courts, prosecutors gions(Western, Southern and Eastern Serbia). and the military. On the other hand, those who are rightwing have more trust in all these institutions. Leftists are less• Support for democracy in general is more pronounced supportive of political parties, institutions of the system and among liberals and those left oriented and less among especially institutions of force, while rightists express signifi- conservatives and right-wing respondents. However, cantly greater trust in them. General trust in people is not conservatives and right-wing respondents are more satrelated to this ideological scale, indicating that it does not isfied with the current state of democracy within the depend on ideology but on other personal characteristics. country. Here we will present the distance of respondents from polit-• Trust in political organizations and institutions is linearly ical party views on the economic left-right scale. The distance related to both the cultural conservative-liberal and ecois expressed in values that can range from 0 to 3.14. The nomic left-right scale. Those who are left and liberal orihigher the value, the bigger the distance of a respondent’s ented tend to express lower trust in political parties, views on Y scales from the party position on the same scale. courts, prosecutors and the military, while those with Respondents are to varying degrees distant from the views of conservative and right-wing attitudes show higher rethe political parties on the economic left-right scale. Re- spect toward these institutions. spondents are on average closest to DJB, NOVA, DS, SRS and are furthest from SNS, SSP, Dveri, DSS and SPS. Observing• The data indicates that the respondents are on average such positioning of the respondents we can see that most of relatively conservative in their views and in that sense, them concentrate around the middle with a moderate ten- they are closer to SNS, SPS and SPAS, the parties which dency towards the left side of the scale. It is interesting that won the elections. It seems that these parties have built the majority of respondents are ideologically closest to those a platform for communication with voters on these ideparties that either chose not to participate in the election or as and it has proved successful. Ideals related to the rule did not pass the census. Among the parties that passed the of law, the independence of institutions, free media and census and achieved good results in the elections, two(SNS the path to Euro integration are topics that resonate and SPS) are quite far from the majority of respondents in with a relative minority of respondents who probably terms of their economic views on this scale, which for sure boycotted the election. can be explained by our sample’s bias towards more educated voters from urban settlements. Based on this data, it• The data also shows that respondents are on average seems that the issues of economic equality, support for the left-leaning on the economic scale left-leaning, emphavulnerable, the role of the state in income distribution are all sizing the significance of social and economic equality, issues of secondary importance. support for the vulnerable, and the role of the state in income distribution. However, the parties that express CONCLUSIONS similar views did not pass the census, while those most distant from these views have won the elections. It • The two KK policy/ideological dimensions are able to seems that the above-mentioned issues are of seconddifferentiate positions of the Serbian public, although to ary importance and that the ideas and policy solutions a modest degree. represented by the parties that won the elections, although contrary to what majority accepts did not affect • Respondents’ views on the economic left-right dimen- the election results. sion are not well differentiated. Most respondents are around the middle or neutral position. • Views on the socio-cultural, conservatism-liberalism dimension are more differentiated. • Respondents’ positions on the two ideological dimensions are weakly associated with gender(women are slightly more liberal and leftist). • The associations with age and education are also low or insignificant. Somewhat higher associations are ob served for the conservative-liberal dimension: more liberal views are associated with relatively higher education and, unexpectedly, older age. 15

ELECTORAL COMPASS 2020 VOTERS’ PREFERENCES IDEOLOGICAL/POLICY DIMENSIONS AND VOTING/VOTER INTENTIONS (PROPENSITY) In this section we check the latent ideological orientation of the voters, expressed through(dis)agreement with 34 statements classified in two dimensions(cultural and economic) against their direct answer to the question about their voting preference. We look at how the voter intentions(vote preferences or vote propensities) are associated with the two ideological dimensions. Note that these analyses are based on a smaller subset of participants, those who were willing to complete the additional questionnaire. Up to 4000 respondents answered the voter probability questions, which is approximately 11-12% of all respondents who participated in the Electoral Compass. This sub-sample is, of course, based on self-selection and personal interest to take part in a more extensive survey. Table 5 shows bivariate correlation coefficients between vote probabilities and respondents’ positions on the two policy dimensions. We can observe statistically significant correla tions for both ideological dimensions. Coefficients in the economic left-right policy column are generally lower and fewer are statistically significant. The highest positive associ ation concerns the voting preference for SNS(r=+.29, p<.001), none of the others approach this level. The highest negative association concerns the voter propensity for the oppositional SSP. But, overall, coefficients are rather low in this column, suggesting a rather small political relevance in the left-right dimension(for similar findings, see for instance Todosijević, 2016). Noticeably higher coefficients are obtained for the conserva tive-liberal dimension. The probability of a vote for SNS and affiliated parties(as well as for some other non-affiliated but conservative, such as DSS and Dveri) is associated with more conservative policy preferences. Relatively more liberal preferences are found among potential voters of parties that represented the opposition at the 2020 elections: SSP, DS, PSG, SDS, NOVA. sion was important for Serbian voters, we would have observed high coefficients between the intention to vote for, for instance a left-wing party and voters’ position on the leftright dimension. In Serbia, the economic left-right dimension is weakly relevant, or completely irrelevant to one’s inclination to vote for most Serbian parties. The second dimension, however, is more politically consequential. Several reasons could be mentioned for the low relevance of the ideological dimension. One is that parties present unclear policy positions on that dimension. For instance, a party can be vague about its position on the left-right dimension. As a result, voters would be unable to make the connection between the party’s position and their own preferences. Also, voters could be uninformed or misunderstand the party’s messages. But it is also possible that voters do not see that dimension as politically salient, even if they are familiar with the positions of the parties. Furthermore, voters could completely disregard ideological dimensions and policy positions of political parties and instead, respond to leaders’ personal characteristics(e.g., charismatic leadership), political scandals, or valence issues(e.g., beliefs that some parties are more or less prone to corruption). It is not clear what particular explanation would fit the Serbi an case--more research is needed, of course. But it is clear that neither parties themselves(as revealed by Electoral Compass experts’ positioning of the parties), nor potential voters take particularly differentiated positions on the economic left-right dimension. If all actors, both parties and voters, take a middle-of-the-road position, then such dimension can hardly guide one’s voter choice decision. In other sections of this document, it is shown that voters don’t always correctly perceive the positions of political parties. So, political education and better information for voters would be welcome. Without adequate information, it is difficult to base the choice on any issue positions. When some issues, policies, or ideological dimensions are important for voter choice, it is revealed in high correlation coefficients. For instance, if the economic left-right dimen 16

ELECTORAL COMPASS 2020 Table 5 Correlations between vote probabilities and policy dimensions Economic LeftRight dimension Probability of vote- SNS Probability of vote- SPATS Probability of vote- SPS Probability of vote- JS Probability of vote- SRS Probability of vote- LSV Probability of vote- DSS Probability of vote- PUPS Probability of vote- DJB Probability of vote- SDS Probability of vote- NOVA Probability of vote- Dveri Probability of vote- NS Probability of vote- PSG Probability of vote- DS Probability of vote- SSP *p<. 0 01. .29* .11* .08* .08* .07* .04 .03 .03 -.03 -.03 -.06* -.07* -.07* -.07* -.09* -.12* ConservativeLiberal dimension -.49* -.25* -.29* -.23* -.39* .14* -.25* -.13* -.04 .21* .22* -.13* .08* .30* .37* .22* The associations between voter probabilities and the cultural conservative-liberal dimension are considerably stronger. Conservative policy preferences are associated with increased probabilities to vote for SNS, SRS and parties close to them (even the Socialist party). On the other side of this dimension are voters from the oppositional block of parties--DS, PSG, SSP, NOVA, SDS... The probability to vote for DJB is the only case of an insignificant association concerning the conservative-liberal dimension. Perhaps, for these voters, some other dimension is relevant, since none of the two included in the Kieskompas model proved such. with small relative differences. Potential voters of the SNS are relatively more right-wing than the average potential DS or SSP voter. 8 A more intuitive picture of the relationships presented by corelation coefficients is provided in Figure 11. It plots the average levels of the two policy dimensions(X and Y) associated with different degrees of propensity to vote for SNS. We can observe, just as the correlation coefficients showed, that this propensity increases as one holds a more right-wing economic position. The slope is considerably steeper for the second, conservative-liberal dimension. There, the tendency is reversed: more liberal positions characterize those who are less likely to vote for SNS. These findings are not surprising. The SNS voters have previously been found to hold a conservative position on life-style issues in general. Still, their economic philosophy hasn’t been very clear thus far. We have to admit that it is not much clearer even with this extensive data. There is little variance in these attitudes among the SNS voters and most of them hold a‘middle of the road’ or perhaps a neutral or undecided position. Whether one intends to vote for SNS or not, is little affected by one’s view of the main economic issues of the day. Note that this is, by far, the strongest association concerning the economic dimension. The probability to vote for other parties is even less associated with economic attitudes. Some previous studies found the association between SNS preference and authoritarian attitudes(Todosijević, Pavlović, & Komar, 2015). Our present finding is clearly in agreement with those results. Figure 11 Probability of vote for SNS and average positions on the policy dimensions One can also observe a kind of parallelism in the associations between the two dimensions. Voter probabilities for parties closer to the right-wing are at the same time higher probabilities to vote for more conservative parties. On the opposite side of the spectrum, voter probabilities for more left-wing parties are accompanied by propensities to vote for culturally more liberal parties. On the one side, this is a frequent finding, given the history of contemporary ideologies, where the left has been associated with emancipatory socio-cultural preferences. But this may also indicate that in Serbia there is really a single political dimension that is relevant for voters, which mixes both economic and cultural issues, but there the cultural issues have a primary role. Note that the presented results mean that the opposition parties(i.e., their potential voters) are particularly left-wing oriented. They are, in fact, not. Voters of all parties are rather close to the center of the left-right dimension. Here we deal There is one additional point worth noting. The major difference, as shown in the graph, is between those respondents who responded“Not likely” and the remaining responses. It seems that although the questions are formulated to represent a continuous scale, there is generally a clear step between points 0 and 1. Those who answered 0 seem to think ‘absolutely not’, while even those who answered 1 implied 8 Average position of SNS potential voters on the economic left-right dimension is at the center(.005), while the average position of DS potential votrers is-.266. 17

ELECTORAL COMPASS 2020 that there is actually a small chance to vote for this party. The implication is that we could recode these variables(voting propensities) into a binary form, which would mean either ‘no chance to vote for a party’, or‘there is some chance to vote for that party’. In the following graph, we present the distribution of the two policy dimensions among those who are certain SNS non-voters and those who would consider voting for SNS. Figure 12 Distribution of the two ideological/policy dimensions and propensity to vote for SNS(Kernel density graph) Figure 13 Distribution of the two ideological/policy dimensions and propensity to vote for SSP(Kernel density graph) Voters of several other parties also show relatively clear profiles on the socio-cultural dimension. Voting propensity for SRS and SPS in one direction and DS and PSG in the other direction is the main example. However, there are parties whose potential voters seem to be completely undifferentiated by their economic philosophy, or their conservatism/liberalism. Voters of DJB and NS are perhaps the best examples. This, of course, does not mean that their voting preferences are not based on some issues. Perhaps, those issues were not included in the KK instrument or are organized differently. The graph shows that although there is a considerably large overlap regarding the distribution of the dimensions among potential voters and non-voters of SNS, the likely voter of this party clearly tends to be more conservative and somewhat more right-wing in matters of economy. The distinction is considerably stronger regarding the socio-cultural dimension. The purpose of showing the same results in three different ways was to provide a more intuitive, and therefore realistic, understanding of the associations between voting propensity and ideological/policy dimensions. Relationships concerning SNS are among the strongest ones, as shown in the table with correlation coefficients. Hence, for the smaller associa tions, the overlaps in distributions, i.e. in policy preferences, are even larger, despite the coefficients being statistically sig nificant. Finally, some parties espouse ideological orientations contrary to their very name and professed orientation. While socialist parties in Europe generally tend to be more progressive in the socio-cultural domain, and leftist in matters of economy, the Serbian Socialist Party is different. First, the propensity to vote for SPS is associated with more conservative views, similarly to SNS. On the other side, the propensity to vote for SPS is associated with a slightly more right-wing economic outlook. Part of the explanation may be the tendency for most parties to rhetorically proclaim leftist economic preferences. In such a context, it may be difficult for a party to distinguish itself as being economically on the left-wing. IDEOLOGICAL DISTANCE AND VOTING PROPENSITY: A MEANINGFUL VOTING CHOICE Another implication that can be read from the correlation table above is that supporters of different parties differ in the degree of their consistency(and distinctiveness) to their economic and socio-cultural policy positions. Parties with voters who take a very consistent and distinctive position on these two dimensions would appear with strong correlations in the above table. What we observed is that the coefficients are not very high and many are rather low. The clearest profile is associated with voting propensity for SNS, where the correlations are the highest for both dimensions. This shows that voting propensities for the other parties are less based on consistent and distinctive policy positions(as defined by the two policy dimensions). Political scientists emphasize that one of the conditions for a meaningful choice(Wessels& Schmitt, 2008) when voting is to recognize party positions on the main ideological dimensions and be able to choose one that is close(r) to the voter. Kieskompas data contains a‘Distance’ variable which represents the Euclidean distance between the respondent’s position and party positions(based on KK expert positioning of the parties), taking into account both ideological/policy dimensions simultaneously. In other words, in the two-dimensional ideological space. When a party and a respondent take the same position on both dimensions, the value of the distance is equal to 0. When they take the opposite positions on both dimensions, the distance is the maximum. This variable is available for all 18

ELECTORAL COMPASS 2020 those who participated in Electoral Compass(not only those who completed the subsequent questionnaire). The range, taking into account all distance variables(i.e. for all parties) is between 0 and 4.35. Figure 14 shows the distribution of the variable Distance to DS(Democratic party). Figure 14 Distribution of the variable distance to DS Figure 15 Relationship between voting propensity and ideological distance, for democratic party(DS) One obvious indicator of‘meaningful choice’ is the association between the distance of the voter’s policy preference and their vote choice(in this case propensity to vote for a party). The greater the distance from a particular party is, the less likely the voter should vote for that party. The smaller the distance, the greater the propensity to vote for that party is to be expected. Thus, a negative association is expected as evidence of‘meaningful choice’. If the association is positive, that would mean that respondents are increasingly likely to vote for that party as it is further from them in terms of the two policy dimensions. That would indicate that policy representation does not work well in Serbian democracy. If the association is close to zero, it would mean that the propensity to vote for that party, positions on the two dimensions are irrelevant. The following graph(Figure 16) shows the same analysis for several parties at the same time. In this way, we can compare the associations more intuitively. We can observe a similar direction and strength of associations for additional parties-PSG, SSP, DjB, next to DS. Thus, the likelihood to vote for these parties is indeed moderately associated with how close ideologically they are to their chosen parties. However, the situation concerning SPAS is different. In this case, the correlation coefficient is positive(although rather low), indicating that the ideological distance from the SPAS party position makes respondents somewhat more likely to give it their vote. Perhaps, respondents like this party for some other reasons and not for their policy positions. Figure 16 Ideological distance and voting propensity: PSG, SSP, DJB, DS, SPAS The results show associations generally in agreement with the hypothesis about the existence of a meaningful choice among Serbian(potential) voters. Overall, the associations between voting propensity and ideological distance are significant but are also of modest magnitude. This means that the distance is important but is not that strict. Figure 15 shows the association between voting propensity and ideological distance concerning the Democratic Party (DS). The slope(predicted scores) shows that the larger the distance, the smaller probability to vote, which is in line with the expectations. Note that the Vote propensity variable is measured on a 0-10 scale. From the graph, we can read that in fact there is a low degree of voting propensity for DS in general(as for all other parties) but that the ideological distance matters. Correlation of r=-.35 is one of the highest observed, although modest in intensity. This means that even when respondents occupy an identical position as the party (DS in this case, i.e., the distance is equal to 0, vote propensity reaches a score of 4, while the range goes up to 10). This is another way to observe the relatively modest relevance of the ideological dimensions for party preferences in Serbia. The next graph(Figure 17) displays another comparative view, now mostly regarding the right-wing parties. We see that for one’s propensity to vote for SNS, ideological distance is important--the larger the distance, the smaller is the chance to vote for SNS. Note that the correlation coefficient(r=-.45) is the highest observed for all parties. The association is weaker but still significant and is also the expected direction for Dveri, SPS and SRS. For JS, policy positions matter little. Again, it is likely that some other, more personal factors play a role in one’s attitude towards this party. 19

ELECTORAL COMPASS 2020 But, the main message is clear: ideological proximity matters most for the likely SNS voters. In other words, likely SNS voters have policy preferences most aligned with their party of choice. It may be due to the fact that the party presents its positions most clearly(in fact, SNS is highly overrepresented in the mainstream Serbian media, which means that they have the means to present their views and reach potential voters). But it may also indicate the difficulties that the oppo sition parties have in establishing their own clear ideological positions and then presenting them to the public ina clear and unambiguous manner. Figure 17 Ideological distance and voting propensity: SNS, Dveri, SPS, SRS, JS sity for SNS and SPS--parties politically opposed to the referent DS. In these cases, the propensity to vote for SNS and SPS increases as the respondents hold more different policy preferences. Indeed, the association between the propensity to vote for SNS and the ideological distance to DS is r=.51--one of the strongest associations obtained in this data set. These ‘discriminating associations’(between vote propensity and distance to competing parties) are generally stronger than associations that show similarity or congruence in policy preferences between respondents and parties. Therefore, it is not just important to have similar policy preferences, it is also relevant to differ from their political opponents. Figure 18 Ideological discrimination: Voting propensity and distance to DS DIFFERENTIATION OF POLICY PROFILES An additional indicator of meaningful choice(Wessels& Schmitt, 2008) is discrimination in terms of party positions on policy dimensions. A party that the respondent does not intend to vote for should on average be more distant from her/ him. Ideologically similar parties should have a similar pattern of associations regarding the policy distance–vote probability relationship. The next several graphs explore the extent to which our respondents are discriminating between parties concerning their policy positions. We expect that the probability to vote for party A will increase as the distance from that party increases if a person is a supporter of party B which is ideologically opposed to party A. Support for ideologically similar parties should reveal a similar pattern of associations for the party respondent him/herself prefers. The relationships observed regarding DS policy positions indicate voter rationality. We expect the same evidence also when the referent case is the policy position of SNS. This is shown in Figure 19. We can observe that greater distance from the policy position of SNS increases vote propensity for the opposed parties--PSG and DS. Also, as expected, this distance is small for those who are more likely to vote for SNS and SRS--parties of similar ideological reputation(although the association is stronger for SNS). The graph also displays the observed association of vote propensity for DJB, which is insignificant. This means that for propensity to vote for DJB it makes little difference how close or distant one is to the policy position of SNS. Obviously, some other issues and considerations are important for DJB(which according to other indicators, not presented here, is a rather unique case). Figure 19 Policy discrimination: Referent party SNS Figure 18 shows that indeed ideological discrimination is in action in Serbia(to some extent). The graph is somewhat challenging for interpretation. Here, the referent party is DS, and all relationships presented are in relation to the distance to DS. We expect that parties that are politically and ideologically closer to DS should exhibit a similar pattern of relationships as, when vote propensity for DS is analyzed. And it is in fact what is observed--vote propensity for SSP reveals the same direction and intensity as vote propensity for DS--the larger the distance, the smaller the vote probability is. The opposite direction of association is observed for vote propen20

ELECTORAL COMPASS 2020 We discuss one additional case of policy discrimination--regarding the distance from the SRS policy position because of one interesting finding. Figure 20 shows that the propensity to vote for ideologically similar parties--SNS and SRS are higher among those holding policy positions similar to SRS. The propensity to vote for parties politically opposed to SRS-DS, PSG, SSP--gets higher the larger policy distance is from the SRS position. The interesting finding is that the slope of the line for SNS is steeper than for SRS. This means that those who are very close to the SRS policy position are actually more likely to vote for SNS than for SRS. The difference is not large but we can observe that it is statistically significant(the confidence intervals around the predicted scores(colored areas for the two lines do not overlap at the low end of the horizontal axis). This could be interpreted in the sense that potential SNS voters really like the SRS policy positions, even though the official SNS policy platform diverged from SRS over years. But, it seems the supporters have not abandoned their ideological roots. Figure 20 Ideological discrimination: Voting propensity and distance to SRS Table 6 shows correlations between vote propensity and positions on specific policy issues, for parties of similar ideolog ical background--SNS, SPS, SRS, and JS. We can observe that positions on many different issues are associated with the propensity to vote for the aforementioned parties. Also, we should note that most coefficients are relatively low in mag nitude. The highest coefficients and highest number of sig nificant coefficients are observed in the propensity to vote for SNS. Only for this party, there are coefficients above.30-7 such instances. Clearly, the propensity to vote for SNS has the strongest basis in specific policy preferences. Particularly strong correlations are observed for issues that could be interpreted as‘the regime issues’, namely those that imply the attitude towards the government(controlled by SNS). The relevant statements refer to whether the government influ ences the work of the judiciary and whether media freedom was threatened during the COVID-19 state of emergency. The next group of high coefficients concerns government evaluation, especially concerning policies related to most recent events--the COVID-19 pandemic and measures introduced by the government. In a way, these issues are not so much ideological but all indirectly refer to the attitude about the ruling SNS party. Some of the other relatively stronger coefficients also concern the evaluation of SNS politics. For instance, SNS vote propensity is associated with the support for state subsidies to foreign investors(r=.34, p<.0001). It is debatable to what extent this attitude reflects a deeper com mitment to specific economic philosophy, or simply the en dorsement of the current government 9 . In fact, when in opposition, SNS criticized the then government for the same kind of policies. We can also conclude that the Electoral Compass respondents demonstrate the basic elements of voter rationality--the ability to base their party preference on party policy positions. Not only are they more likely to vote for parties with congruent policy profiles, but they are also less likely to vote for parties with contrary policy profiles. VOTING INTENTIONS(PROPENSITY) AND SPECIFIC POLICY PREFERENCES In the previous section, we saw that the general Kieskompas policy dimensions are associated with party preferences in Serbia--moderately but generally as expected. Now, we look into how specific issues are associated with vote propensity. It is difficult to summarize the results since the dataset covers 16 parties and 34 specific issues. A simple table of correlation coefficients would contain 544 entries. Moreover, each of the relationships could be examined in multiple ways and by grouping parties and issues into different categories. Given the amount of space such detailed analyses would take, we will show here only a fraction of the main findings. Vote propensities for the remaining three parties reflect what was observed for SNS, but in a weaker, paler form. The coefficients are in almost all cases in the same direction but they are generally of lower magnitude. So, for instance, they all tend to think that“Russia is a more reliable partner for Serbia than the EU”, just as supporters of SNS do. They are also more in support of regime issues(not surprising since SPS and JS are part of the ruling coalition, while SRS is a kind of ‘friendly opposition’). One interesting deviation concerns the issue of Serbia joining the EU. A moderately negative attitude about this issue is observed concerning the vote propensities of SRS, SPS, and JS, but no association in the case of SNS. It seems that SNS supporters adjusted to the official policy of the party, though just to the degree as not to be opposed to the EU. On the other side, supporters of the associated, ideologically similar SRS, SPS and JS, can still cultivate their negative sentiments about joining the EU. 9 The same would apply, for instance, to the issue of the dual education. 21

ELECTORAL COMPASS 2020 Table 6 Coefficients of correlation between voting propensity and positions on specific policy issues SNS SPS SRS JS Membership in the EU would be useful for Serbia.-.10-.28-.11 To maintain independence, Serbia should never join NATO..12.09.12.07 Russia is a more reliable partner for Serbia than the EU..28.21.29.17 Serbia needs to recognize Kosovo’s independence.-.12-.11-.21-.10 Serbia should respect Montenegro’s right to regulate laws on religious communities.-.15-.12-.22-.11 Same-sex marriages should never be allowed in Serbia..14.12.13.09 Serbian government needs to implement laws that preserve patriarchal family values..21.17.22.15 Serbia should ban the settlement of migrants coming from non-European countries..08.08.08.06 In Serbia, the government essentially influences the work of the judiciary.-.53-.23-.25-.17 Giving more autonomy to the regions leads to the further disintegration of Serbia..20.14.23.11 Serbian citizens should be allowed to vote for individuals...-.12 Churches and religious communities must pay taxes.-.16-.13-.29-.16 Subsidies to state-owned companies that are losers should be abolished,...-.14-.10-.17-.09 Protecting the environment is more important than economic growth.-.17-.09 Dual education serves nothing but providing cheap labor.-.24-.11-.07 Serbia should encourage the arrival of immigrant workers.-.06 Everyone is responsible for finding a job for themselves,... To attract foreign investors, the state should provide subsidies....34.16.15.14 The minimum wage should be increased, even if it reduces the company’s profit.-.05 Debt collection should be the responsibility of the state, not private agencies.-.11 The communal police should be abolished because they are of no use.-.21-.08-.11-.07 To keep young professionals, the state should donate apartments to them..07.11.10.13 Pensions can only be increased if economic growth allows it..10 All adult citizens of Serbia should receive 100 euros each....21.13.11.11 The health care system is neglected and more needs to be invested in it.-.32-.13-.16-.11 The new circumstances are an opportunity for the domestic economy and tourism to recover- our goods and services should be favored. .19.15.13.11 Citizens’ health is a priority at the time of the Corona,....12.08 Measures taken by the state to help vulnerable companies and people who have lost their.47.20.16.12 jobs are quite sufficient for recovery. During a state of emergency, it is necessary to listen to the state regardless of the possible.45.23.20.16 consequences for civil and political freedoms. During the pandemic, China showed that it is Serbia’s greatest friend..44.26.27.20 The measures applied to pensioners during the pandemic were overly restrictive...-.43-.20-.22-.14 These elections need to be postponed...-.53-.28-.30-.19 Media freedom was threatened during the state of emergency.-.61-.28-.28-.20 Pollution in Serbia occurs because citizens drive too much cars and heat with wood and coal..07 Note: Only statistically significant coefficients are included in the table(p<.001). Finally, it is worth noting some instances where correlations are insignificant or particularly low. For instance, it seems that environmental issues do not provoke any consistent reactions among supporters or opponents of these parties. The relevant coefficients are low or insignificant. Also, several economic issues are low or insignificant as well. Issues such as whether the state should help with unemployment, minimum wage and pensions are all basically unrelated to respondents’ intention to vote for these parties. Perhaps, particularly interesting is that the policies concerning immigrants are almost completely unrelated to party preferences. It is surprising since voters of these parties are known to endorse more ethnocentric, prejudiced views. Radical right-wing parties(such as SRS, out of whom SNS originated) are almost by definition nationalistic and anti-immigrant. Even the SRS vote propensity is barely correlated with policies referring to migrants. This may be a consequence of SNS leadership refraining from anti-immigrant rhetoric(in which they excelled before coming to power) and their supporters picked up the clue. It could also indicate that the issue was successfully kept out of public focus. The COVID-19 pandemic and the atmosphere surrounding it certainly contributed to such issues losing salience at the time of the election and this survey. 22

ELECTORAL COMPASS 2020 Table 7 Coefficients of correlation between voting propensity and positions on specific policy issues- the opposition parties DS PSG SSP SDS Membership in the EU would be useful for Serbia. .30.23.16.18 To maintain independence, Serbia should never join NATO. -.18-.15-.10-.13 Russia is a more reliable partner for Serbia than the EU. -.27-.21-.14-.15 Serbia needs to recognize Kosovo’s independence. .24.20.10.17 Serbia should respect Montenegro’s right to regulate laws on religious communities. .22.15.08.14 Same-sex marriages should never be allowed in Serbia. -.21-.15-.13-.12 Serbian government needs to implement laws that preserve patriarchal family values. -.23-.16-.10-.14 Serbia should ban settlement of migrants coming from non-European countries. -.20-.13-.08-.14 In Serbia, the government essentially influences the work of the judiciary. .13.17.09 Giving more autonomy to the regions leads to the further disintegration of Serbia. -.20-.17-.10-.15 Serbian citizens should be allowed to vote for individuals... .06.06 Churches and religious communities must pay taxes. .10.13 Subsidies to state-owned companies that are losers should be abolished,... .07 Protecting the environment is more important than economic growth. Dual education serves nothing but providing cheap labor. .06 Serbia should encourage the arrival of immigrant workers. .16.08.06.12 Everyone is responsible for finding a job for themselves,... To attract foreign investors, the state should provide subsidies... The minimum wage should be increased, even if it reduces the company’s profit. Debt collection should be the responsibility of the state, not private agencies. The communal police should be abolished because they are of no use. -.07 To keep young professionals, the state should donate apartments to them. .06 Pensions can only be increased if economic growth allows it. All adult citizens of Serbia should receive 100 euros each... -.15-.07-.08-.09 The health care system is neglected and more needs to be invested in it. .07.07.06 The new circumstances are an opportunity for the domestic economy and tourism to recover -.24-.11-.14-.12 our goods and services should be favored. Citizens’ health is a priority at the time of the Corona,... Measures taken by the state to help vulnerable companies and people who have lost their jobs -.19-.15-.17-.09 are quite sufficient for recovery. During a state of emergency, it is necessary to listen to the state regardless of the possible con-.19-.15-.16-.08 sequences for civil and political freedoms. During the pandemic, China showed that it is Serbia’s greatest friend. -.27-.19-.18-.12 The measures applied to pensioners during the pandemic were overly restrictive... .21.17.17.09 These elections need to be postponed... .24.22.20.11 Media freedom was threatened during the state of emergency. .27.25.22.16 Pollution in Serbia occurs because citizens drive too much cars and heat with wood and coal. -.06 Note: Only statistically significant coefficients are included in the table(p<.001). NOVA .15 -.15 -.17 .18 .17 -.17 -.19 -.17 -.13 .08 .14 -.09 -.06 -.10 -.16 -.10 -.08 -.15 .10 .07 .14 To summarize, for support of SNS and related parties, the The overall impression is that the preferences are more based following issues are important: on current political considerations than on deeper ideological commitments. • the regime issue • current policy evaluations Table 7 shows correlations between vote propensity and po• symbolic issues(Russia, Kosovo, China). sitions on specific policy issues for the main opposition par ties--DS, PSG, SSP, SDS and NOVA. In this case, there are Somewhat unexpectedly, it turned out that irrelevant are fewer significant coefficients and those that are significant these issues: are of lower magnitude. For parties of this orientation, the regime issues are also relevant, but the coefficients, expect • environmental protection edly, go in the opposite direction. So, for instance, three of • economy(redistribution) these parties have potential voters who are more inclined to • policy towards immigrants. 23

ELECTORAL COMPASS 2020 think that the government is affecting the judiciary in Serbia. They all find that media freedom was endangered during the Corona crisis and that the elections should have been postponed. Evaluations of the current policies are also relevant, but this time negative evaluations. The inclination to vote for these parties is associated with a critical view of the government’s• policy concerning pensioners, various COVID-19 policies such as the“100 Euro policy” 10 and similar. Preferences for these parties seem to show a more ideological character, as well as a strong basis in symbolic issues. Potential voters of these parties are more pro-EU and pro-NATO orient-• ed(well, less negatively oriented towards NATO), less enthusiastic about Russia and China and more respectful of“Montenegro’s right to regulate laws on religious communities.” The symbolic issues, relating to life-style matters and general tolerance is also present here. For instance, supporters of these parties are less negative about immigrants, less negative about same-sex marriages and less enthusiastic about• patriarchal family values. However, neither this bloc of parties is particularly profiled in the realm of economy and environment. There are at least four explicitly economic issues where none of the vote propensities appear as a significant correlate. Thus, the two blocs of parties may be fiercely opposed politically, but that seems to be entirely unrelated to their views of the economy. Likewise, the two environmental issues are irrelevant for vote propensities for any of the five parties presented in this table.• CONCLUSIONS • The two Kieskompas policy/ideological dimensions are able to differentiate positions of the Serbian voters, although to a modest degree. The overall impression is that the quality of democracy in Serbia would benefit from more political education and informative discussion about the policy positions of different parties. • The economic left-right dimension is particularly poorly associated with vote intentions. The exception is a moderate association of the likelihood to vote for SNS and right-wing leaning. All the remaining coefficients are considerably lower or insignificant. However, there are some clear tendencies. Respondents who intend to vote for parties ideologically and politically associated with the ruling SNS party tend to be slightly more on the• right-wing, while the opposition voters tend to be slightly more to the left-wing compared to the average respondent. • The conservative-progressive dimension is much more associated with voting preferences in Serbia. The pattern of the associations, however, mirrors those ob10 Helicopter money as a measure of support to citizens. 24 served for the left-right dimension. SNS and the associated parties, including SRS, plus some other ideologically more conservative parties(DSS, Dveri) are relatively more on the conservative side. Voter intention for the opposition parties is associated with more liberal stances. Potential voters of DjB represent an exception in this picture. Their views are not associated with either of the two dimensions. It is possible that their views are not well differentiated, but also that their distinctive ideological outlook is not captured by these two dimensions. The analyses showed that most of the Serbian(intended) voters pass the‘voter rationality test’. Namely, it proved that the ideological distance to a party is associated with voting propensity. In other words, Serbian voters are more likely to vote for parties that are closer to them ideologically, or policy-wise than for more distant parties. Although the associations are not of impressive strength (for DS the correlation coefficient is r=-.35, and for SNS r=-.45; these are the highest coefficients observed), they indicate that democratic representation is not absent from Serbian electoral politics. Voters tend to support parties with more similar policy profiles. This does not apply equally to voters of all parties. Some of the associations are rather low or insignificant--for instance, in the case of JS and SPAS voters. A somewhat more demanding test is to not vote for parties that differ from one’s policy preferences. This requires the voters to be able to differentiate between parties who have similar positions from those occupying different policy positions for themselves. It proved that, indeed, vote propensity declines as parties have more different policy positions from potential voters. For instance, respondents are more likely to vote for SNS or SPS because their policy preferences are more different from those of DS. Likewise, voters are more likely to vote for DS and PSG the further their policy preferences are from those of SNS. So, yes, elements of rational, policy-based voting intentions are recorded by the Electoral Compass. Although the degree is not too impressive, nor it applies to all parties, it demonstrates the basic democratic competence of Serbian voters. However, there is a lot of room for improvement. Positions on specific issues(i.e. not on the general di mensions that summarize the individual issues) are also associated with vote propensities. SNS proved to have the clearest policy profile, as there is the highest number of significant and relatively high coefficients for their in tended voters. They probably received a clearer message about the position of this party. From one angle, this is expected, as the media is strongly dominated by the content related to this party. The opposition parties have much more difficulty accessing the media and therefore promoting their policy positions.

ELECTORAL COMPASS 2020 • It should be emphasized that the strongest associations,• We would like to emphasize two implications of the preboth for the governing parties and the opposition, relate sented findings. One is the need for political education to the‘regime cleavage’ theme. These would be items and more substantial information for the public about that express positive or negative views of the current party policy positions. These are requirement for develregime(controlled by the SNS). The highest coefficients oping a more functional democracy in Serbia. More refer to whether the government influences the work of open and politically balanced media is important here. the judiciary and whether media freedom was threat- But other efforts aimed at drawing public attention toened during the COVID-19 state of emergency. The oth- wards policy positions of different parties, such as Elecer relatively strong associations either express evaluation toral Compass, can have an influential role. In an envi of specific policies implemented by the government, or ronment where there is more demand for information concern‘symbolic’ issues, such as the attitude towards about party policy positions, parties themselves are likeRussia and the EU. The classic ideological issue concern- ly to provide more clear and unambiguous information ing economic, redistributive issues is poorly represented, about their actual positions. while the ecological issues are completely unrelated to party preferences.• The second important implication is that the political spectrum needs to be better differentiated. While the • As was the case with the general dimensions, the coef- Serbian party system has many parties, in addition to ficients associated with preferences for the opposition the currently dominant party, it is clear that they are just parties are also relatively weaker(and insignificant more separated into two blocs--parties associated with the often). But the pattern of the relationships pretty con- government(and SNS), and those opposed to them. sistently reflects the associations with the governing There is very little differentiation within both camps. At parties. One distinction seems to be that in the case of least according to the issues examined by the Electoral these parties’ voters, the symbolic issues have a some- Compass. Even the relatively new parties, such as DJB, what stronger relative importance--especially the EU is- NOVA, or SPAS, failed to demonstrate clear and distincsues, the Kosovo issue and same-sex marriages. So, it tive policy profiles to their potential voters. It is certainly seems that potential voters of the liberal opposition are possible that the study omitted issues relevant to these differentiated by the‘Eurocentric’ and libertarian out- parties. However, the study did include the main issues look, in addition to the dislike of the governing parties. that were discussed during the election campaign. Again, this implies the need for both: the abilities of parties to present their policy positions and the need to encourage the public to take those positions into account when making their vote choice. 25

ELECTORAL COMPASS 2020 REFERENCES Barberá, P., Bauer, P. C., Ackermann, K.,& Venetz, A. (2017). Is the left-right scale a valid measure of ideology? Individual-level variation in associations with“ left“ and“ right“ and left-right self-placement. Political Behavior, 39(3), 553-583. Caughey, D., O’Grady, T. O. M.,& Warshaw, C. (2019). Policy ideology in European mass publics, 1981–2016. American Political Science Review, 113(3), 674-693. Cochrane, C. (2013). The asymmetrical structure of left/right disagreement: Left-wing coherence and right-wing fragmentation in comparative party policy. Party Politics, 19(1), 104-121. Elster, J., Claus, O., Preuss, U. (1998) Institutional Design in Post-communist Societies, Cambridge University Press, Cambridge, UK. Kuzmanović, B. (2010). Autoritarnost – vapaj za jakim i pouzdanim vođama i disciplinom. In Grupa autora, Kako građani Srbije vide tranziciju(str. 87-104). Beograd: Friedrich Ebert Stiftung, Centar za studije socijalne demokratije i CeSID. Mihailović, S. (1991). Izbori’90: mnjenje građana Srbije. U: S. Mihailović i V. Goati(ur.), Od izbornih rituala do slobodnih izbora: sondaža javnog mnjenja uoči prvih višestranačkih izbora u Srbiji(str. 27−126). Beograd: Institut društvenih nauka. Oesch, D.,& Rennwald, L. (2018). Electoral competition in Europe‘s new tripolar political space: Class voting for the left, centre‐right and radical right. European Journal of Political Research, 57(4), 783807. Pantić, D. (2003). Kulturno-vrednosni rascepi kao determinante partijskog pregrupisavanja u Srbiji(str. 95-128). In J. Komšić, D. Pantić i Z. Slavujević, Osnovne linije partijskih podela i mogući pravci političkog pregrupisavanja u Srbiji. Beograd: Institut društvenih nauka. Pavlović, Z.& Todosijević, B. (2016). Ideological self-identification and party identification in Serbia. In: Bondarevskaya, Irina& Iordănescu, Eu gen(eds.), Psychology of Political and Economic Self-Constitution(pp. 58-62). Institute of Social and Political Psychology, NAES of Ukraine. Pavlović, Z., Todosijević, B.& Komar, O. (2019). Education, Authoritarianism, and Party Preference in the Balkans. Problems of Post-Communism, 66:6, 402-417. Spasojević, D. i Stojiljković, Z. (2020). Između uverenja i interesa. Beograd: Fabrika knjiga. Stojiljković, Z. (2009), Partijski sistem Srbije. Beograd: Službeni glasnik. Stojiljković, Z. , ed.(2012) Kako glasam, zašto glasam, ako uopšte glasam. Beograd: FES. Todosijević, B. (2013). Serbia. In S. Berglund, J. Ekman, K. Deegan-Krause& T. Knutsen(eds.), The Handbook of Political Change in Eas tern Europe, Third Edition(str. 523-565). Cheltenham: Edward Elgar. Todosijević, B. (2016). Left-Right ideology: Its meaning and effects on party preferences in Serbia. Sociološki pregled, Vol. 5, No. 2, pp. 161-178. Todosijević, B., Pavlović, Z.& Komar, O. (2015). Obrazovanje, ideologija i politika: Značaj razlika u nivou obrazovanja za političke podele u Srbiji i Crnoj Gori.[Education, ideology and politics: The significance of education cleavage for political divisions in Serbia and Montenegro.] Srpska politička misao, Posebno izdanje, 2015, pp. 125-152. Wessels, B.,& Schmitt, H. (2008). Meaningful choices, political supply, and institutional effectiveness. Electoral Studies, 27(1), 19-30. 26

IMPRESSUM AUTHORS IMPRESSUM Dr. Slobodan Cvejic is professor of sociology at the University of Belgrade. He has been dealing with social research for more than 30 years. He also took part in numerous international research projects and provided consultancy to international organizations and government agencies. Dr. Dušan Spasojević is an associate professor at the Faculty of Political Sciences, University of Belgrade. His main research areas are political parties, social cleavages theory, civil society, populism and the process of transition in post-communist societies. Dr. Dragan Stanojević is assistant professor at the Faculty of Philosophy, Department of Sociology. His key qualifications are: 12 years of doing social research in Ser bia and the SEE region as well as participation in national and international research projects. His research and policy work is focused on family relations, youth and children, education, social inequalities, social inclusion and life course patterns. Dr. Bojan Todosijević is a Senior research fellow at the Centre for Political Studies and Public Opinion Research, Institute of Social Sciences(IDN) in Belgrade, Serbia. His research focuses on understanding social and political attitudes, and political behavior. His publications dealt with topics such as the structure of social attitudes, the influence of ideological orientations on political attitudes and behavior, nationalist attitudes, political intolerance, and populist ideology. Friedrich-Ebert-Stiftung| Office in Belgrade Dositejeva 51| 11000 Belgrade| Serbia Phone+381 11 3283 285 info@fes-serbia.org www.fes-serbia.org Heinrich Böll Stiftung| Office in Belgrade Kapetan-Mišina 25|11 000 Belgrade| Serbia Phone:+381 11 400 59 77 info@rs.boell.org www.rs.boell.org ComDmieeirncidaileusseer Pouf balilkmateiodniazpuumblAisuhseddrubcykthgeebFraiecdhrtiechn-AEbnesircth-Stteinftusinngd nicht (FESn)oatnwdeHnedingreicrwh eBiösell dStieiftduenrgFr(HieBdSri)cihs-nEobterpte-Srmtifitutendg.wDitiehsoeuPt uthbelikwatriot-n ten wcoirndseanutf oPfapthieerFaEuSsannadchHhBaSlt.iger Forstwirtschaft gedruckt.

ELECTORAL COMPASS 2020 Analysis of political landscape in Serbia The current political spectrum in Serbia is insufficiently differentiated. While the Serbian party system has many parties, in addition to the one currently dominant party, it is clear that they are just separated into two blocs--parties associated with the government(SNS), and those opposed to them. There is very little differentiation within both camps. SNS proved to have the clearest policy profile. Their intended voters probably received a clearer message about the positions of this party than the tentative voters of oppositional parties. In order to develop a more functional democracy in Serbia we need organized political education and more substantial information for the public about party policy positions. More open and politically balanced media is important for this but the political parties are also tasked to make their political messages clearer.