Party Responsiveness to Public Opinion in Young Democracies

Are political parties in young democracies responsive to the policy preferences of the public? Compared to extensive scholarship on party responsiveness in established democracies, research on party responsiveness in young democracies is limited. We argue that weaker programmatic party–voter linkages in post-communist democracies create incentives for parties to respond to their supporters rather than the more general electorate. Such responsiveness occurs in two ways. First, parties follow shifts in the mean position of their supporters. Second, drawing on the research on party–voter congruence, we argue that parties adjust their policy positions to eliminate previous incongruence between themselves and their supporters. Analyses based on a comprehensive dataset that uses expert surveys, parties’ manifestoes and election surveys to measure parties’ positions, and several cross-national and national surveys to measure voters’ preferences provide strong support for this argument.


Introduction
The 2015 parliamentary election in Poland brought a victory to Law and Justice (Prawo i Sprawiedliwość, PiS), the main opposition party. The party's election campaign was characterised by its criticism of the liberal economic policies of the incumbent government and harsh anti-immigrant rhetoric. The party's strategy was 'clearly about firing up the base, not about appealing to moderate voters' (Tworzecki, 2019: 103-104).
Whether such patterns of party responsiveness to the public's policy preferences are widespread in young democracies is relatively unknown. Compared to the large body of scholarship on party responsiveness in established democracies (e.g. Adams et al., 2006;Ezrow et al., 2011), research on party responsiveness in new democracies is limited despite the importance of party responsiveness for democratic representation. The present study addresses this omission by asking the following question: are political parties in young democracies responsive to the policy preferences of the public?
We argue that established parties (defined as parties that existed as significant electoral forces for at least two consecutive general elections) in young democracies primarily respond to their partisan supporters. They do so in two ways. First, parties follow shifts in the mean position of their supporters. Second, parties also adjust their policy positions in order to eliminate previous incongruence between themselves and their supporters. Our argument draws on and combines insights from two contrasting perspectives on party democracy in post-communist and other young democracies.
On one hand, high levels of electoral volatility and the electoral success of new parties (Haughton and Deegan-Krause, 2015;Tavits, 2008), the prevalence of antiestablishment parties with vague programmatic appeals (Sikk, 2012) and the less significant role of the left-right dimension  provide indirect indications that parties' responsiveness to public opinion is weak. We draw on this body of research by arguing that "Central and Eastern European (CEE)" parties do not respond to the general electorate. On the other hand, 'programmaticness' of party politics in the region is suggested by relatively programmatic appeals of political parties (Benoit and Laver, 2006), stability of the structures of competition (Rovny and Polk, 2017) and moderately high congruence between parties' and their supporters' ideological positions (Kitschelt et al., 1999). We build on this work by suggesting that parties respond to their partisan supporters.
Weaker programmatic positional and competence-based party-voter linkages lead parties in post-communist democracies to respond not to the general electorate but to their supporters. Compared to their Western European counterparts, CEE electorates are less likely to identify the substantive content of the broader ideological dimension , to locate their own and parties' positions on this dimension  and to use this information when casting their votes (Burlacu and Tóka, 2014). They are also more disappointed with the performance of established parties (Pop-Eleches, 2010;Roberts, 2010).
However, there are significant differences between partisan supporters and independent voters. Partisans in CEE democracies are characterised by higher levels of ideological structuration and voting and satisfaction with the performance of established parties. CEE parties are therefore able to find policy positions that respond to partisan supporters' preferences on multiple issues and can reasonably expect to be electorally rewarded for this responsiveness.
Meanwhile, most independent voters are strongly disaffected by the performance of established parties, lack structure in their policy preferences or are characterised by weak ideological voting. Established parties are therefore unable to develop policy packages that are responsive to the concerns of many independent voters and/or lack electoral incentives to do so. Overall, established parties primarily respond to their partisan supporters.
Our empirical analyses are based on the most comprehensive dataset on parties' and voters' left-right positions ever assembled in the context of Central and Eastern Europe. Specifically, we combine data on parties' positions from the Chapel Hill Expert Survey (CHES), Manifesto Project Database (MARPOR) and a set of national and European election surveys with data on the public's preferences from the same set of election surveys as well as the European Social Survey (ESS). While time periods covered by each of these sources vary, as a whole the dataset considers two and a half decades of democratic party competition in Central and Eastern Europe. Our analyses show consistently strong support for our hypotheses despite the diversity of data sources and the complications these introduce for research on party positional shifts (cf. Adams et al., 2019). We therefore contribute to the literature on party competition and representation in young democracies, which often emphasises the lack of party responsiveness to public opinion.
With regard to the broader debate on party representation and party spatial competition, we make two important theoretical contributions to the understanding of party responsiveness. Work on Western Europe (WE) has shown that party characteristics (goals, organisation or ideological breadth) influence which voter groups parties respond to (Bischof and Wagner, 2020;Ezrow et al., 2011;Schumacher et al., 2013). We argue that the strength of programmatic linkages between parties and voters also has an important role in the existence and type of party policy responsiveness. Importantly, this concerns not only the extent to which citizens vote based on their policy and ideological positions but also voters' evaluations of party competence. Widespread dissatisfaction among voters leads parties towards responding to their partisan supporters rather than the general electorate.
Second, our argument combines two distinct traditions in the study of party representation. The first one focuses on the responsiveness of parties to the shifts in voters' preferences. We build on this research by suggesting that parties in CEE democracies follow shifts in their supporters' preferences. The second strand of research examines static party-voter congruence. We draw on this work by hypothesising that parties also seek to reduce or eliminate previously existing incongruence between their and their supporters' policy positions. The innovative combination of these two distinct but complementary hypotheses provides a more complete understanding of party responsiveness to public opinion in young democracies.

Party Responsiveness in Established Democracies
In contrast to the substantial body of research on the responsiveness of parties' policy positions to public opinion in WE, comparable scholarship focusing on CEE democracies is largely absent. 1 As a starting point in developing our argument, we examine the Western European literature, which suggests two types of party responsiveness. First, parties have responded to shifts in the centre of the distribution of the whole electorate's policy preferences (e.g. Adams et al., 2004Adams et al., , 2006Ezrow et al., 2011;but see Meyer, 2013, andO'Grady andAbou-Chadi, 2019, for contrasting evidence). 2 This hypothesis is mainly derived from a set of multiparty spatial models that assume vote-seeking parties and probabilistic voting behaviour (Lin et al., 1999). In addition, office-seeking parties respond to the central voter tendency in order to be sufficiently attractive in the formation of government coalitions (Lehrer, 2012).
Other scholars argue that parties are more responsive to shifts in the preferences of their supporters (the mean party voter). This is particularly the case for niche parties because they prioritise policy goals, are dominated by policy-seeking activists or have narrow issue appeals (Bischof and Wagner, 2020;Lehrer, 2012;Schumacher et al., 2013).
Our theoretical argument, as developed in the next four sections, not only builds on these contributions but also draws on the research on party-citizen linkages, new political parties and party-voter ideological congruence. In line with most research on party responsiveness, we focus on established parties (parties that existed as significant electoral forces for at least two consecutive general elections) that obtain the bulk of electoral support, despite high electoral instability in CEE democracies. 3

Positional Party-Voter Linkages in Young Democracies
Programmatic linkages between parties and voters develop when parties compete for voter support by providing them either policy packages (positional linkages) or competence to address problems that voters consider important (competence-based linkages) (Kitschelt, 2007). In this and the next section, we show that in CEE countries both types of linkages are relatively strong for partisan supporters but weaker for most independent voters. This leads parties to be policy-responsive to only their supporters.
The programmatic positional linkage is based on parties making clear, salient and distinct offers on policy issues or dimensions, and voters having informed preferences on these issues and dimensions, knowing which party best represents these preferences and voting on the basis of these perceptions Kitschelt, 2007). Since voters are often unaware of parties' specific policy positions (and new issues constantly arise), parties also need to bundle their issue stances into positions on a broad ideological dimension, such as the left-right dimension. The left-right dimension then provides a useful heuristic that helps citizens to orient themselves in politics even without knowing parties' positions on most issues. The left-right framework is therefore often used to summarise voters' and parties' policy preferences in cross-national comparative work, including that on post-communist democracies Kitschelt et al., 1999;Savage, 2016). We employ the same approach here.
Empirical evidence suggests that, in CEE countries, programmatic voting on the leftright dimension is characteristic mostly to partisan supporters. Starting with country-level means, CEE electorates are less ideological than those in established democracies. Voters in CEE democracies are less likely to place themselves on the left-right scale and their left-right positions are less anchored in substantive issue positions Noël and Thérien, 2008;Rohrschneider and Whitefield, 2012), their placements of parties on the left-right spectrum are more heterogeneous and less precise (Best and McDonald, 2011; and the effect of voters' left-right placements on vote choice is also weaker (Burlacu and Tóka, 2014;Dalton, 2011).
However, these averages hide substantial variation at the individual voter level. In line with classical arguments (Converse, 1964), education, political sophistication and interest increase the levels of voter self-placement on the left-right dimension (Zechmeister, 2015), more educated voters err less when placing parties on the left-right scale (Marinova, 2016) and political sophistication and education have a strong effect on the probability of casting an ideological vote (Dalton, 2011;Kroh, 2009). By finding parties that represent their interests, these more educated, sophisticated or interested voters are more likely to become partisans (Brader and Tucker, 2001;Lupu, 2015;Miller et al., 2000) in line with the rationalist theories of partisanship (Franklin and Jackson, 1983). 4 Thus, while in established democracies ideological left-right voting is widespread despite individual-level differences, the lower overall levels of ideological voting in CEE democracies mean that it is mostly partisan supporters -voters with relatively high levels of education, political sophistication or interest -that are sufficiently ideological and open to programmatic positional appeals from parties.
To strengthen this point further, Table 1 presents information on several well-established indicators (cf. Best andMcDonald, 2011, andWhitefield, 2012, of programmatic positional party-citizen linkages in CEE and WE). Despite the diversity of measures, they all suggest that citizens in European democracies can be divided into three groups in terms of the strength of the programmatic positional linkages. WE partisans score highest on all indicators, followed by the WE independents and CEE partisans with moderately high scores. The last group is the CEE independents, whose levels of programmatic structuration and voting are low and at least one standard deviation below the WE independents and CEE partisans.

Competence-Based Party-Voter Linkages in Young Democracies
CEE democracies are also characterised by weak linkages related to politicians' perceived competence at delivering good government performance. Compared to established democracies, voters' perceptions of the competence of established parties are considerably lower in CEE democracies (Roberts, 2010: 70). This reflects both objective factors Voter is coded as relatively close on the left-right scale to the party he or she voted if voter's self-placement is 0-3 and mean placement of the party by all voters is lower than 4.5 or voter's self-placement is 4-6 and mean placement is 3.5-6.5 or voter's self-placement is 7-10 and mean placement of the party is higher than 6.5 (cf. Best and McDonald, 2011: 97). e Online Appendix 2 provides the list of issues.
f Average score based on the survey question using a 0-10 scale.
(relatively high levels of corruption, painful economic transitions, the loss of communistera social protection nets) and expectations among voters of a quick catch-up with levels of prosperity in WE (Holmes, 1997). However, these regional means again hide substantial individual-level differences between partisans and independents. Even in CEE countries, some voters have relatively high evaluations of party competence, leading them to form partisan ties in line with the rationalist theories of partisanship (Brader and Tucker, 2001;Fiorina, 1981). The reverse relationship -partisans projecting competence on their parties -further increases the correlation between partisanship and competence attributions (Campbell et al., 1960). Furthermore, as discussed above, partisans are more likely to engage in programmatic positional voting than are independent voters; the resulting higher proximity between their preferences and parties' positions among the partisans also increases their perceptions of party competence (Rohrschneider and Schmitt-Beck, 2016). Table 1 strengthens this point by presenting three indicators of competence-based linkages in Europe: the share of partisan and independent voters who consider all established parties as not being able to deal with the issue most important to the voter, satisfaction with government performance and satisfaction with the performance of the party that the respondent voted for. WE partisans are most satisfied with the performance of established parties; WE independents and CEE partisans follow with moderate levels of satisfaction, while CEE independents stand out as particularly disappointed with the competence and performance of established parties.
The weakness of competence-based linkages among the independent voters accounts for the almost universal electoral punishment of incumbent parties that occurs in CEE democracies even under favourable economic conditions (Roberts, 2010). In the first few democratic elections, voter disaffection benefitted established opposition parties. However, after trying both mainstream centre-left and centre-right parties, in later elections dissatisfied voters have increasingly turned towards new parties that emphasised their superior competence or employed anti-elite and anti-corruption rhetoric (Engler, 2016;Hanley and Sikk, 2016;Pop-Eleches, 2010;Tavits, 2008). Yet, the frequent emergence of new parties has not strengthened the competence-based linkages, and many voters supported them primarily as a protest against established party alternatives (Pop-Eleches, 2010). Even where voters expected better performance from new parties, these expectations were quickly disappointed once they entered governmental office, often leading the disappointed voters to support yet newer parties (Haughton and Deegan-Krause, 2015). Low competence attributions to all parties have also been linked to lower electoral turnout rates (Rohrschneider and Schmitt-Beck, 2016), suggesting that voters may prefer to abstain instead of casting a vote for any of the established parties that they consider incompetent.

Parties' Strategy: Responsiveness to Partisan Supporters
The discussion in the previous two sections can be summarised as follows. Most partisans in CEE democracies are fairly ideological and moderately satisfied with the competence of established parties -less so than their Western European counterparts, but to a similar extent as the independents in WE. In contrast, most CEE independents are largely nonideological and/or have negative perceptions of the performance and competence of established parties. Thus, CEE independents frequently support new parties or abstain in elections.
How do these characteristics of party-voter linkages shape policy responsiveness of CEE parties to public opinion? Following other authors (Kselman et al., 2016), we assume that, at least partially due to their weak societal roots and lower levels of intra-party democracy, CEE parties are predominantly vote-and office-seekers. They will therefore select responsiveness strategies that maximise their electoral support and increase the chances of entering government coalitions with other parties.
We argue vote-and office-seeking CEE parties primarily respond to the preferences of their partisan supporters. This is because they have (1) capability and (2) electoral incentives to do so, and (3) this strategy is compatible with parties' office-seeking incentives.
First, relatively high levels of structuration in partisans' policy preferences increase parties' capability to develop policy positions that bring them close to their partisans on the broad ideological dimension and multiple policy issues. Meanwhile, the issue profiles of most independent voters are diverse. This presents a difficult challenge for an established party that aims to find positions that bring it relatively close to this segment of the electorate on multiple policy issues. At best, an established party would be able to respond on the broad ideological dimension to a minority of the independents whose preferences are more structured. 5 Second, established parties have vote-seeking incentives to respond to their partisans who are more policy-conscious and should reward the party's efforts to address their policy concerns. Partisans also tend to have better opinions about the competence of their party and will support it even if another party is equally proximate to their policy preferences.
On the contrary, established parties lack electoral incentives to be policy-responsive to most independent voters. Even if some citizens in this group have sufficiently structured preferences for parties to be able to respond to them, they may not be able to identify parties' policy positions and/or would not vote based on policy concerns. A yet smaller group of independent voters who do have structured preferences and vote based on policy concerns would not reward the efforts of an established party to respond to their policy preferences due to their disaffection with the competence of established political alternatives; instead, these more ideological independents are likely to support a new party or abstain from voting.
Responsiveness to their supporters is also an electoral strategy that is largely compatible with parties' office-seeking goals. Policy and ideological proximity has limited effects on the formation of government coalitions in post-communist democracies. In the early years of democratic regimes, this was a result of the regime divide between communist successor and other parties (Grzymala-Busse, 2001), but the same pattern has persisted in later years (Savage, 2016). The limited role of policy concerns in government formation means that, from an individual party's perspective, adopting noncentrist positions that are generally in line with the preferences of its supporters (Rohrschneider and Whitefield, 2012) does not reduce its chances of entering government. The trade-off between office-and vote-seeking strategies is particularly alleviated by the frequent entry of electorally successful but less programmatic new parties that attract the disappointed and less ideological voters. Specifically, as new parties often enter government coalitions (Savage, 2016), thus providing coalition partners for the more programmatic established parties, the latter need to make fewer policy concessions to acquire governmental office.
We also note that our argument, while using WE as a baseline for understanding the electoral context in CEE democracies, does not imply that (mainstream) parties in WE respond to the general electorate. Higher ideological structuration and voting in WE among both independents and partisans provides more favourable conditions for parties in this region to respond to the mean voter, but there may be other reasons for them (e.g. increasing party system fragmentation) to focus on their supporters instead. Indeed, as argued above, empirical evidence on the question remains mixed. 6

The Relationship between Congruence and Responsiveness
How do parties respond to their supporters? In their recent review of the research on ideological congruence, Golder and Ferland (2018) argue that scholars of party responsiveness should consider the role of party-voter congruence. They suggest that parties respond not only to the shifts in voters' positions (as per arguments in the research on party responsiveness) but also to the previous incongruence between themselves and voters. This is because the congruence between parties' and voters' positions is the 'ultimate goal' which can be achieved through responsiveness (Golder and Ferland, 2018: 3). Empirically, congruence between parties and their voters is indeed high Rohrschneider and Whitefield, 2012).
Following these arguments, we theorise that parties respond to two types of information about their supporters' preferences. First, a party has strong incentives to eliminate any initial incongruence (resulting from, for example, interest group influence, misguided electoral strategies or international or economic constraints on a party's policy) by moving towards its supporters' position. Thus, if the average party supporter is initially to the right of the party's position, and remains stable, the party would move to the right to remove this incongruence. Second, parties follow shifts in the policy positions of their supporters, thus helping the party to avoid a situation whereby the distance between its policy position and that of its supporters (incongruence) increases. For example, if the party was congruent with its supporters some time ago, but then its supporters move to the right, then an equivalent move by the party assures that the levels of congruence remain high. 7 The two types of responsiveness jointly determine the total size of the shift in a party's policy position. For example, if the average party supporter is initially somewhat to the right of the party's position, and moves further to the right, the party would move substantially to the right to address both previously existing incongruence and the further shift in its supporters' preferences. If, however, the average supporter position is initially somewhat to the right but subsequently moves to the left, the party would not change its position as it would likely be congruent with its supporters because of their shift.
Based on these arguments, we formulate two hypotheses on party responsiveness in CEE democracies:

Data and Measurement
To provide a systematic examination of our hypotheses, we rely on information about parties' and voters' positions on the left-right dimension in 10 Central and Eastern European countries (all European Union (EU) member states in the region except Croatia) that are arguably the most successful democracies in the post-communist region.
While the Manifesto Project Database (MARPOR; Budge et al., 2001;Klingemann et al., 2006) has been most widely used for studying shifts in parties' policy in WE, we complement its data with parties' positions on the left-right dimension derived from the CHES conducted in 2002-2014 and election surveys (European Election Study, Comparative Study of Electoral Systems and national election studies; Bakker et al., 2020;Comparative Study of Electoral Systems, 2018;Schmitt et al., 2009Schmitt et al., , 2016van Egmond et al., 2014). This helps us make more reliable inferences given that all three methods have strengths and weaknesses (Benoit and Laver, 2006).
Using MARPOR data, we construct two measures of the change in parties' left-right positions between consecutive elections. The first one relies on the RILE scale, the most commonly used measure of left-right ideology (e.g., Budge et al. 2001). However, due to the specific content of the left-right dimension in CEE countries (Benoit and Laver, 2006: 202), the RILE scale may be less applicable in this context (Mölder, 2016). We therefore use the estimates of the Bayesian item response theory (IRT) model fit to the MARPOR data by Däubler and Benoit (2017) as a second measure. The model produces estimates of how each category in the MARPOR project relates to the left-right dimension without relying on a pre-existing scale and has been shown to correlate more strongly with the experts' placements of parties' positions than the RILE scale (Däubler and Benoit, 2017). In both cases, we compute the difference in the left-right positions of parties in two consecutive elections. Since, as discussed below, the Party Supporter Gap Hypothesis requires us to compute the difference between the mean party supporter position and the party's position, we rescaled left-right positions before computing the change variable. Specifically, the RILE scale of parties' positions, which originally ranges from −100 to 100, was recoded to the 0-10 scale used in voter surveys. The Bayesian IRT estimates were rescaled to a standardised variable with the mean of 0 and standard deviation of 1. Standardisation was conducted within each country to account for potential cross-national differences in the positioning of political parties.
From the CHES dataset, we use the question in which experts are asked to place parties on the left-right dimension in terms of their overall ideological stances. The question uses an 11-item scale, in which 0 and 10 represent the most leftist and rightist positions, respectively. We use the mean expert placement as an estimate of the party's position and compute the change in the party's policy position as the difference between its left-right positions in two consecutive CHES waves.
From the election surveys, we use the questions in which respondents were asked to place individual parties on a 0 to 10 scale for the left-right dimension. Where an original scale was different, it was recoded to the 11-item scale. A party's policy position is estimated as a mean placement of it by the 40% top educated respondents. This follows research suggesting that less well-informed voters tend to place parties in centrist positions (Golder and Stramski, 2010). Where available, we used Comparative Study of Electoral Systems (CSES) or national election surveys from two consecutive elections. 8 In several cases, these were complemented by the surveys from the European Election Study conducted close to the national election years. The shift in a party's policy position is computed by taking the difference between its positions in each pair of surveys.
To measure the preferences of party supporters, we draw on two sets of surveys. The first combines European Election Study (EES), CSES and national election surveys that have also provided one source of information on parties' positions. We create three separate datasets by combining the variables on voters' preferences from election surveys with each of the three sources of information on parties' positions. Second, we also use the ESS (2020) in combination with the CHES expert surveys. The ESS matches the timing of the CHES expert surveys better than the election surveys and, unlike the latter, is a single cross-national survey. The list of countries and years covered by each data source on parties' and voters' positions is listed in Online Appendix 3.
Voters' policy positions are estimated based on the question in which they were asked to place themselves on the left-right scale. Where an original scale was not an 11-item scale, it was recoded to such. The measures based on this question therefore by definition exclude voters who could not place themselves on the left-right dimension. The exclusion of respondents unwilling to indicate their left-right placements is in line with the theoretical hypotheses that suggested party responsiveness to their partisan, more ideological supporters.
We operationalise partisan supporters based on the questions on respondents' closeness to parties. Online Appendix 4 provides more details on the questions used.
To test the Party Supporter Change Hypothesis, measures showing the shift in the mean partisan supporter position were coded in each survey dataset. Similarly, for the Party Supporter Gap Hypothesis, the variable capturing the difference between the positions of the mean partisan identifier and the party is constructed in each dataset. While the same 0-10 scale was used in voter surveys and expert surveys, for the datasets that combine party manifestoes and voter surveys the left-right scales are different. As mentioned above, the RILE scale was recoded to the 0-10 scale. The Bayesian IRT estimates of parties' positions and mean supporter positions on the left-right scale were standardised (within each country) to the variables with the mean of 0 and standard deviation of 1.
As a key control variable, we use the change in the mean independent voter position. 9 As discussed in the theoretical section, while this variable was shown to have an effect on party policy change in WE, we have strong theoretical reasons to expect that it will not matter in the CEE context. Two additional dichotomous variables indicate party transformations (splits, mergers and electoral coalitions -common in the context of young democracies) that should lead to leftist and rightist change in party's position. These party transformations are common in CEE countries. They may lead to shifts in a party's policy position if, for example, a substantial faction with different ideological preferences than the rest of the party leaves or if a party joins an electoral alliance or merges with another party that represents a different ideology. To account for these changes, we identified major splits (characterised by at least 10% of a party's legislative delegation joining the splinter party), mergers and electoral coalitions based on Ibenskas and Sikk (2017). We then constructed two dichotomous variables based on whether a party transformation should lead to a leftist or rightist change in its position. Thus, leftist positional changes are expected if (1) a party experienced a split and the splinter faction was to the right of the rump party, (2) if the party left an electoral coalition with another, more rightist party or (3) if the party joined an electoral coalition or merged with a more leftist party than itself. Reverse coding was applied to the expected rightist shifts. Table 2 Table 3 reports the results of statistical analyses testing the Party Supporter Change and Party Supporter Gap Hypotheses. In total, five linear regression models have been fit, one for each combination of the measures of parties' and voters' positions. All models control for country-specific effects through the use of dichotomous variables for countries. Our data are nested within parties and country-time period dyads (for the datasets  Poland 2002Poland -2006 or elections (for the datasets that use manifestoes or election surveys to measure parties' positions). We include random intercept terms for each of these levels.

Results
The results of statistical analysis largely support the Party Supporter Change Hypothesis. Change in the mean partisan supporter has a statistically significant effect (at the 0.01 level) in Model 1 (expert surveys used to measure parties' positions) and Model 5 (voter surveys used to measure parties' positions). The substantive size of these effects is quite large: a one point change in the mean party supporter position changes party's positions by "0.4 points (Model 1) 0.5 points (Model 5).
Models 3 and 4 that use manifesto data to measure parties' positions are also positive, but not statistically significant. This could be a result of methodological issues related to the use of MARPOR data for measuring parties' policy positions (Benoit and Laver, 2006), particularly in the CEE context where parties' manifestoes may be a less reliable data source on parties' policy (Däubler and Benoit, 2017: 13). Also, parties may find it more challenging to reflect very recent changes in their supporters' preferences in their manifestoes. In contrast, parties are better placed to respond to such shifts in their speeches and media appearances, both of which influence the perceptions of parties' positions held by experts and voters.  The results in Table 3 demonstrate even stronger support for the Party Supporter Gap Hypothesis. Strikingly, the distance between the mean supporter position and their party's policy position has a statistically significant effect (at the 0.01 or lower level) in all five statistical models despite the use of a variety of data sources for parties' and voters' positions. The direction of the effect is as hypothesised. Thus, if the position of the mean party supporter is to the right of their party's position at the beginning of the inter-election period or the time period between two consecutive CHES waves, the party's position moves to the right by the end of the period. Conversely, a party's policy shifts to the opposite direction if the mean supporter is placed to the left of it. The substantive size of the change in a party's position varies from 0.1 points on the RILE scale in the party manifesto dataset, 0.3 to 0.4 points in the expert survey dataset and 0.9 points in parties' leftright positions derived from election surveys.
Substantively weaker effects in the manifesto dataset are in line with the results on the Party Supporter Change Hypothesis, which was supported with expert and mass survey but not party manifesto data. This might be a reflection of lower importance of party manifestoes in CEE party politics and/or methodological challenges of the Manifesto Project Database in the context of this region. However, unlike the change in the mean party supporter position, the incongruence variable remains statistically significant using the manifesto data. This may suggest that parties find it easier to adapt their manifestoes to close the 'gap' that existed a few years ago (as suggested by the Party Supporter Gap Hypothesis) as opposed to responding to more recent changes in the policy preferences of their supporters (as implied by the Party Supporter Change Hypothesis).
In Online Appendix 5, we provide a robustness check by testing our hypotheses using vote choice questions (as opposed to partisanship questions used in Table 3). Thus, the Party Supporter Change Hypothesis is tested using the shift in the mean position of voters indicating that they supported the party in the most recent national election or intended to support the party in the next general election. The Party Supporter Change Hypothesis is tested using variables capturing the difference between the positions of the mean party voter and the party. The results are very similar to those in Table 3. Furthermore, while in the main analyses we do not control for previous shifts in party policy positions (i.e. the lagged dependent variable), our results are not changed when this variable is included, even if the sample size is reduced quite significantly (see Online Appendix 6).
In additional models (not reported in the article), we included as controls party electoral support, ideological centrality, participation in government, the change in economic growth between the end and beginning of each time period considered (cf. Ezrow and Hellwig, 2014) and disproportionality of electoral system (measured by natural log of average district magnitude). None of these variables had a statistically significant effect on the change in parties' positions in more than one out of five datasets combining different measures on parties' and voters' positions. The inclusion of these variables also had a very limited impact on the effect of the two variables of substantive interest. We therefore left out these controls from the final set of models. We also examined whether the effect of the variables testing our two hypotheses varies with electoral institutions, party's government status and party electoral support. Analyses with these interaction effects show no clear evidence that parties' responsiveness depends on either of these three variables.
One concern with the empirical test of the Party Supporter Change Hypothesis is that, despite strong theoretical grounding, it might be driven by the partisan sorting process (Adams et al., 2011) in which voters respond to parties' policy shifts by either adjusting their policy positions (as a result of persuasion by the party) or party choice/affiliation. However, if such partisan sorting takes place, at the aggregate level it should be preceded by a change in average supporter perceptions of parties' policy positions that in turn follow prior shifts in actual parties' positions. Additional statistical analyses in Online Appendix 7 show no effect of the shift in party policy positions on the change in the perceptions of parties' positions by the mean party voter and mean partisan supporter. This suggests that the correlation between the shifts in parties' and voters' left-right positions is driven by the responsiveness of parties, not the type of partisan sorting process described by Adams et al. (2011). This evidence is consistent with findings from WE showing that, while parties' policy shifts have only weak effects on voters' perceptions of parties' positions, parties nevertheless respond to the shifts in voters' preferences because this might bring marginal electoral benefits, especially if the parties send a consistent message (Adams, 2012). Furthermore, it is likely -and should be explored in future research -that voters' perceptions and behaviour depend on previous incongruence between voters' and parties' policy positions (rather than shifts in parties' positions). In this case, in line with our argument, parties would have electoral incentives to stay congruent with their supporters by both removing previous incongruence and responding to shifts in voters' preferences.
The example of the Slovak Democratic and Christian Union-Democratic Party (Slovenská demokratická a kresťanská únia -Demokratická strana, SDKÚ-DS) illustrates these findings. In the early 2000s, the party supported economically liberal measures to revive the Slovak economy, thus finding itself to the right of its supporters, particularly on economic issues (3.8 points on the 0-10 redistribution scale based on CHES and ESS data; the gap on the general left-right dimension was 0.5 points). In response to this gap, in 2008 the party put forward a new programme that included support for housing development and an increase in maternity benefits, while in 2009 it supported the increase in the duration of unemployment benefits to contain the social consequences of the economic crisis (SITA Slovak News Agency, 2009). These measures were aimed at, in the language of party leadership, the 'middle classes and young families' -the core electorate of the party that consisted of younger, urban and more educated voters (SITA Slovak News Agency, 2009). These shifts are consistent with the more general strategies of centre-right (and other) parties in Slovakia who 'often paid too much attention to fighting with each other for the "core" reformist right electorate, and too little attention to attracting middle-of-the-road floating voters' (Henderson, 2012: 2). In the 2010 election, this rather narrow electoral focus worked to the benefit of the SDKÚ-DS: it mobilised its core electorate, remained the most significant centre-right force with 15% of the vote and also achieved its office goals due to the electoral success of the new antiestablishment party 'Freedom and Solidarity' that joined the government led by the SDKÚ-DS.
Besides the support to our two hypotheses, another important result of our analyses is the absence of support for the idea that parties in CEE countries respond to the broader electorate beyond their own supporters. In all models in Table 3, the coefficients of the variables measuring mean independent voter change are negative, although not statistically significant. Similarly, the effect of the change in the mean voter position is also not statistically significant, as we report in Online Appendix 6. Thus, in contrast to substantial research showing that mainstream Western European parties respond to the general electorate (Adams et al., 2006;Ezrow et al., 2011;Schumacher et al., 2013), this type of responsiveness seems less relevant in post-communist democracies.
As we argue in the theoretical section, this is likely because a substantial share of the independent voters have diffuse policy preferences without much ideological structure, do not recognise parties' positions or, even if they do, do not engage in ideological voting. This makes it difficult for parties to identify policy positions that could represent these voters and, even if they succeed in doing so, risky to rely on these voters to notice and reward this effort. Moreover, many independent voters are disappointed with the competence of all established parties. Policy responsiveness to these groups is unlikely to bring electoral benefits. This leaves the responsiveness to a party's own supporters as an optimal strategy. This strategy is also in line with parties' office goals.
Another control variable -party transformations -also have no effect on party policy change. The only case in which this variable is significant is in Model 1, in which rightist transformations surprisingly lead to leftist change in parties' policy positions. This result, however, may be driven by a small number of observations (five parties experienced rightist transformations). We reran the analyses excluding the two party transformation variables as a robustness check. The results are substantively similar to those presented in the main analyses.

Discussion
This research presents the first systematic study of party responsiveness on the left-right dimension in Central and Eastern Europe over a relatively large time span in these newer democracies. Combining information on parties' positions from their manifestoes, expert surveys and voter surveys, and two separate survey datasets on voters' preferences, we report two key findings. First, established parties in CEE countries only respond to their supporters. We argue that this is a result of the nature of party-voter linkages in CEE democracies. The majority of independent voters are less ideological and/or disappointed with the performance of established parties, which reduces the capability of and incentives for established parties to respond to the broader electorate. This makes policy responsiveness to supporters an electorally advantageous strategy. Second, we show that CEE parties, aiming to be aligned with their supporters, respond not only to the dynamic shifts in their preferences but also to the prior incongruence between themselves and their supporters.
Evidence of party responsiveness to their supporters provides an important contribution to the literature on party politics in post-communist democracies, in which high electoral volatility, party instability, voter disappointment with the performance of democracy and low electoral participation are frequently considered as indications of weak representation. Our argument and empirical findings nuance this perspective. Established parties respond to the partisans, but much less so to the independent voters who readily switch between parties or support new parties. Furthermore, our finding that CEE parties respond to their supporters by trying to stay congruent with them underscores the importance of analysing congruence and responsiveness as two dimensions of party representation. As such, these results reveal, at least partially, the process through which the high levels of left-right congruence between CEE parties and their supporters found in previous research (Rohrschneider and Whitefield, 2012) is achieved.
Nevertheless, our findings also raise a number of important questions for future research. While we find that those parties that survive longer than a single election on average respond to their supporters, there might be factors that condition this responsiveness. For example, parties with narrow issue appeals could have lower incentives to be responsive on broad programmatic dimensions while parties with less developed organisations might be less capable of doing so. The extent to which parties are responsive with regard to actual policies they implement while in office is also an important question (cf. Roberts, 2010). Last but not least, the extent to which voters perceive and respond to the shifts in parties' ideological appeals -only briefly addressed here -is another important area for future research.
Our findings also have implications for the broader literature on party competition and representation. The strength of the responsiveness of CEE parties to prior incongruence between themselves and their supporters raises the question of the prevalence of such responsiveness elsewhere as part of the broader emerging research agenda on the relationship between congruence and responsiveness (Golder and Ferland, 2018). Furthermore, while research on party representation in WE suggests party characteristics (goals, organisation and the broadness of issue appeals) as explanations of the type of party responsiveness (Bischof and Wagner, 2020), we show that broader electoral and party contexts can be equally if not more important factors. Future research could consider the importance of party-voter linkages in a broader cross-regional comparative study.

Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: J.P. recognises gratefully funding from the Swedish Research Council (Vetenskapsrådet), grant number 2016-01810.

Supplementary Information
Additional Supplementary Information may be found with the online version of this article.

Contents
Online Appendix 1: Responsiveness in the pursuit of congruence Online Appendix 2: Issue-specific questions in Table 1 Online Appendix 3: Datasets on parties' and voters' policy positions Table 1: CHES + ESS dataset Table 2: CHES + EES/CSES/NES dataset Table 3: MARPOR + EES/CSES/NES dataset Table 4: EES/CSES/NES dataset Online Appendix 4: Survey questions on voting and partisanship Online Appendix 5: Comparison of the results for party voters and partisan supporters Table 5: Comparison of results for party voters and partisan supporters Online Appendix 6: Public preferences and party policy shifts with lagged party policy change Table 6: Public preferences and party policy shifts (with lagged party policy change) Online Appendix 7: Party responsiveness and partisan sorting Table 7: Supporters' perceptions of parties' policy shifts Notes 1. But see, for example, Rohrschneider and Whitefield (2012), Hanley and Sikk (2016) and Tavits (2013) for work on representation in the region. 2. In the subsequent theoretical and empirical analyses, in line with most empirical literature on the effect of public opinion on party policy, we use the mean as a statistic of the central point in the voter distribution. 3. On average, established parties (including mergers and coalitions of previously existing parties) receive 80% of the votes in parliamentary elections in CEE countries (Engler, 2016;Tavits, 2008). 4. Based on European Social Survey (ESS) Waves 1-8 data, in the CEE countries analysed here, the share of partisans among respondents with less than secondary education, secondary education, post-secondary non-tertiary and tertiary education were 27%, 34%, 36% and 43%, respectively. 5. However, a party can still reach out to independent voters using other strategies, such as personal appeal of party's leadership or candidates. 6. We also note that our argument is different from , who show that lower share of partisans in post-communist democracies means that parties with more extreme policy positions are more successful electorally if voter preferences are polarised. We argue that parties respond primarily to partisan supporters (who can have moderate or extreme preferences) because they are not capable and lack electoral incentives to respond to most independent voters. 7. Ferland (2018) also examines the idea -in the Western European context -that parties should respond to both initial incongruence and subsequent public opinion shifts, but develops different hypotheses than ours. Online Appendix 1 compares his hypotheses with ours and justifies our approach. 8. Two exceptions are the Czech Republic 1996-2002and Slovakia 2010 in both cases, there was one election in the middle of the period covered for which election survey (ES) data were not available (1998 for the Czech Republic and 2012 for Slovakia). 9. This variable was standardised in the analyses that use the item response theory (IRT) estimates of parties' positions. 10. Due to space constraints, the descriptives for party transformations are not included. The share of observations experiencing either kind of transformation is 13% (CHES (Chapel Hill Expert Survey) + ESS), 11% (CHES + ES) and 21% (all other datasets).