Who Responds? Voters, Parties, and Issue Attention

Who Responds? Voters, Parties, and Issue Attention Heike Klüver 1 University of Bamberg heike.kluever@uni-bamberg.de Jae-Jae Spoon University of North Texas spoon@unt.edu ABSTRACT: Do parties listen to their voters? In this article, we address this important question in the study of democratic representation by exploring whether parties respond to the issue priorities of voters. We move beyond looking at position congruence to explore how citizens issue concerns influence the policy issues parties emphasize. We argue that political parties respond to voters, but that their issues responsiveness varies considerably across different party types. More specifically, we hypothesize that large parties are more responsive to voters policy priorities while government parties listen less to voters issue demands. In addition, we expect that niche parties are not generally more responsive to voter demands, but that they are more responsive to the concerns of their supporters in the policy issue areas they own. To test our theoretical expectations, we combine data on the issue attention of 113 parties from the Comparative Manifestos Project with data on the policy priorities of voters from the Comparative Study of Electoral Systems and various national election studies across 18 European democracies in 63 elections from 1972 until 2011. Our findings have important implications for understanding political representation and democratic linkage. Paper prepared for presentation at the Elections, Public Opinion and Parties Conference Lancaster, UK, 13 15 September 2013 1 The order of authorsʹ names reflects the principle of rotation. Both authors have contributed equally to all work. 1

1. Introduction One of the central functions that parties have in democracies is to link citizens with political decision makers (Lawson 1980, 2005; Dalton et al. 2005). In recent years, however, there has been concern about whether parties still perform this important role in the face of increasing partisan dealignment, expanding mass media, and growing interest group activity (Dalton and Wattenberg 2000). Importantly, for parties to stay relevant in democracies in an era of challenges from other actors and institutions, they must be responsive to voters demands. Indeed, Stimson et al. (1995) developed the concept of dynamic representation to describe how a party shifts its policy position in response to changes in public opinion. Parties which have not been responsive to voters issue concerns can lose voters to newly formed parties. Kitschelt (1988, 209), for example, argued that left libertarian parties are likely to form when the unresponsiveness of existing political institutions coincides with favorable political opportunities. The formation of European green parties is often attributed to the interaction of these two conditions (see e.g. Markovits and Gorski 1993). More recently, we have seen the development of the Pirate Parties in many countries, which is largely a result of other parties failure to respond to voters policy concerns regarding information privacy, transparency, and sharing. However, not all parties are equally responsive. Parties that are more focused on policy (Strøm and Müller 1999; De Swann 1973; Wittman 1973, 1983), for example, may be less responsive to the demands of the electorate as a whole than they are to those of their own party supporters. Adams et al (2006), for example, found that mainstream parties responded to public opinion; whereas, niche parties did not. Similarly, in their research, Ezrow et al. (2011) show that mainstream parties respond to shifts in the mean voter position; whereas niche parties respond to shifts in the mean position of their own voters. Much of the current research on party responsiveness to voters has focused on preference congruence. Specifically, this research has examined how parties policy positions are influenced by changes in voter preferences (e.g. Adams et al. 2004, 2006, 2009; Ezrow and Hellwig 2011). In other words, if the electorate shifts to the right, then the expectation is that parties will move 2

their positions to the right. While this research has significantly advanced our understanding of the relationship between parties and voters, it has not addressed how parties respond to the issue priorities of voters. A second area of research, which has focused on the relationship between the policy priorities of citizens and government policy has, moreover, sought to address how voters issue preferences affect government policy (see, e.g., Wlezien 1995; Stimson et al. 1995; Jones and Baumgartner 2004; Hobolt and Klemmensen 2008; Jennings and John 2009; Chaqués Bonafont and Palau 2011; Bevan and Jennings 2013). In this article, we seek to bridge these research areas. Current research has offered some insight into what determines the attention that parties pay to political issues. Parties policy priorities are based on the issues they own (Budge and Farlie 1983; Petrocik 1996); they are placed on the agenda by issue entrepreneurs or other parties in the system (e.g. Green Pedersen and Mortensen 2010; Hobolt and De Vries 2011); and they are affected by the political and economic context (Duverger 1954; Riker 1982; Inglehart 1997). In line with these previous findings, we posit that political parties not only respond to shifts in voter preferences, but that they also listen to voters when deciding which policy issues to emphasize. Thus, similar to Spoon and Klüver (N.d.), we argue that there is a positive relationship between the attention that voters pay to a given policy issue and the attention that parties devote to the issue. However, while Spoon and Klüver (N.d) demonstrate that the electoral context has an important impact on the degree of issue responsiveness of political parties, they focus less on the ways in which parties characteristics could affect party responsiveness. In this article, we recognize that there are important differences among parties and that parties responsiveness is therefore conditioned by several party features namely size, type, and governing status which is our central argument. We thus seek to understand not only when parties are more responsive to the issue priorities of voters, but which parties are more responsive. Our findings have important implications for understanding the nature of political representation across parties in Europe. 3

To understand the relationship between parties and voters and how it differs across parties, we examine parties responsiveness to both the electorate s and their own voters issue priorities from 1972 to 2011 in 18 European countries. More specifically, we analyze the responsiveness of 130 parties across 13 issue areas in 63 national elections. We measure party issue attention using the data from the Comparative Manifestos Project (CMP). Issue priorities of the entire electorate and of party supporters are measured using data on the most important issue/problem which we obtained from the Comparative Study of Electoral Systems and various national election studies. This paper proceeds as follows. First, we present our theoretical argument about party responsiveness in general and how it is conditioned by party characteristics and derive hypotheses that guide the empirical analysis. In the next section, we elaborate on our data and methods, including how we created our issue categories. Next, we present the results of our empirical analysis, which seeks to understand what influences which issues parties choose to emphasize in their manifestos and how this varies across different types of parties. Finally, we conclude by discussing the implications of our findings for political representation. 2. Parties and Issue Responsiveness In this section, we present out theoretical argument in detail. We begin by outlining our assumptions about party behavior from which we derive the argument that parties are responsive to the issue priorities of voters. We then present our argument for why responsiveness differs across parties. 2.1. Issue responsiveness of political parties To understand why parties may or may not be responsive to the policy priorities of voters, we need to understand what drives their behavior. Political parties are typically conceptualized as rational, goal oriented and purposeful collective actors that aim to maximize the achievement of their preferences (Downs 1957). We assume that parties may pursue two different goals. First, 4

Riker (1962, 1982) suggested that parties are primarily office seeking actors which seek to gain control of executive offices. Accordingly, parties seek to win elections for instrumental reasons as they are primarily interested in obtaining control of political offices. Winning elections means first and foremost controlling the executive branch, or as much of it as possible. Second, De Swaan (1973) argued in contrast that parties are policy seeking actors. According to this party behavior model, political parties value certain policy goals and their actions are primarily motivated by these policy objectives. As Strøm and Müller (1999) have pointed out, these goals are not necessarily independent of each other. Political parties might instrumentally strive for offices in order to use government responsibility to implement their preferred policies. Similarly, political parties might instrumentally seek to achieve policy goals in order to attract voters to ultimately gain control over executive offices. It is therefore not surprising that research on party competition has been characterized by a long standing debate about which of these goals most adequately captures party behavior (see e.g. Strøm 1990; Strøm & Müller 1999). We thus do not assume that either one of these goals solely explains party behavior. Instead, we posit that both goals affect party behavior, but that there are important differences with regard to party type. Some parties value office goals over policy goals while other parties primarily act as policy seekers. To win office, parties need to attract as many voters as possible. However, changing policy positions may be difficult to achieve as parties are committed to certain positions due to their ideology or their reputation (see e.g. Robertson 1976; Kollman, Miller and Page 1992). Thus, another important and promising instrument through which parties can gain votes is the attention they pay to different political issues. We therefore assume that rational, goal oriented, parties will seek to emphasize the issues that voters prioritize as means to win political office. They will, moreover, adjust their policy priorities to reflect the changing priorities of voters. (Downs 1957). We build on the saliency theory of party competition and the issue ownership model (Budge and Farlie 1983, Petrocik 1996), and argue that, to win elections, parties will focus on the issues voters emphasize. Instead of highlighting the same (or all) issues, both of these theories suggest that parties compete with each other by emphasizing different policy issues. 5

Parties will select the issues on which they have an advantage. Thus, a left wing party may choose to focus on labor issues; whereas a right wing party may choose to focus on nationalism. When voters change their issue priorities, we expect parties to follow. Just as voters are often retrospective (Fiorina 1981; Kiewiet and Rivers 1984; Powell and Whitten 1993), we argue that parties also act retrospectively. Adams and Somer Topcu (2009) have demonstrated that parties gain votes in the current election when they have moderated their votes in the previous election; thus, there is a lagged effect of responsiveness. 2 Building on this finding, we argue that parties need time to process voters demands and thus will respond in the subsequent election. Spoon and Klüver (N.d.) accordingly also show that parties respond to voters issue priorities from the previous election. Thus, our first hypothesis is as follows: H1: The higher the attention that voters pay to a given policy issue at t 1, the higher the attention that parties pay to this issue at t0. 2.2. Are some parties more responsive? Parties vary in several important ways among these differences are size, governing status, ideological breadth, and issue focus. These different characteristics, moreover, have been shown to influence the goals that parties have and the decisions that they make. Parties behavior is motivated by emphasizing different goals, and depending on the goal that is emphasized, behavior may vary. For example, if a party primarily focuses on policy, it may choose to only focus on a specific set of issues in the election campaign or choose to stay out of government if participating may mean too much compromise of its core policies; such has been the case with the Swedish Green Party (Bale and Bergman 2006). 2 Following the extant literature, we recognize the reciprocal relationship between voters and parties, that parties are not only responding to voters preferences, but that voters might also respond to parties issue emphasis (e.g. Adams et al. 2004, 2009; Bélanger and Meguid 2008). In this study, however, we choose to focus only on party responsiveness to voters and save unpacking the reciprocal relationship for future research. 6

Compared to smaller parties, mainstream or large parties tend to be office seeking and follow a broad ideological appeal; whereas small parties are often more ideologically consistent and are more focused on policy (Downs 1957; Kirchheimer 1966; Harmel and Janda 1994; Cox 1997). Large parties are often referred to as catch all parties and seek to bring as many voters into their tent as they can (Kirchheimer 1966). Over time, these parties may change their ideological position or issue focus to attract more voters (Przeworski and Sprague 1986; Harmel and Janda 1994), as the British Conservative Party has done under David Cameron s leadership. Conversely, smaller parties size is often due to the very choices they have made regarding policy positions and issue priorities. They may have explicitly decided to not appeal to the broader electorate, but to focus instead on a smaller set of issues, which may not have as wide of an appeal. As a result of these core differences, scholars have found that niche or smaller parties behave differently than larger parties in election campaigns (Adams et al. 2006; Ezrow et al. 2011; West and Spoon 2013) and in government (Bäck and Dumont 2006; Bolleyer 2007; Heller 2002; Jensen and Spoon 2010). For example, Adams et al. (2006) and Ezrow et al. (2011) demonstrated that small or niche parties are less responsive to the mean voter and Heller (2002) found that regional parties in Spain chose to stay out of the national government in exchange for policy concessions from the governing party. Following this research, we argue that larger parties will be more responsive to voters than small parties, which is our second hypothesis. H2: Mainstream or large parties will be more responsive to voters than small parties. Next, we consider the responsiveness of governing parties. Although governing parties are often larger parties, smaller parties have also been included in governing coalitions, such as the green parties in Belgium, Finland, Germany, and Italy. When parties obtain office, we argue that their status changes. While larger parties are more likely to be responsive than smaller parties, once a party is in the government, it may be more constrained in its actions compared to opposition parties, regardless of its size. As governing parties are focused on policy making, they may have less ability to compromise their policy priorities. Political parties that are in government at the time of an election have been responsible for policies that have been 7

implemented in the past years. They have been exposed to intensive scrutiny by the media, which monitor governmental action and communicated legislative decisions to voters. As a result, voters have fairly good knowledge about the policy decisions of government parties. It is thus very difficult for a governing party to entirely change its policy priorities as a response to shifts in voter attention to policy issues. For instance, imagine a government that has largely ignored environmental policy during its four year legislative term. Even though voters might care a lot about environmental protection, it would not be very convincing to emphasize this issue in the election campaign since the government has not done anything about that issue while being in office. A governing party may therefore compromise its credibility among voters if it changes its issue priorities right before an election (for a similar argument, see also Nanou and Dorussen 2013). 3 In contrast, opposition parties have much more flexibility to propose policies that the voters may favor more. They have not been in office for a while and past activities in government may have faded in voters memory. Opposition parties are not as constrained by outside forces such as economic downturns or international commitments and can typically make policy proposals throughout the election campaign without having to refer to their legislative decisions. For this reason, political parties that are in opposition at the time of the election can be more responsive to the policy priorities of voters. They, moreover, have the incentive to be more responsive to regain control of government. We therefore expect that governing parties are less responsive to changes in the issue emphasis of voters than opposition parties, which are not as constrained by their own past decisions. Opposition parties can more easily adjust their policy priorities without having to suffer credibility losses. 4 Our third hypothesis is thus the following: 3 One could imagine, however, how an exogenous shock right before an election could motivate a governing party to be more responsive to voters issue priorities. 4 Governing parties inability to change their policy priorities may also be explained by theories of issue ownership and competency. Because they were elected on issues that they own (Budge and Farlie 1983; Petrocik 1996) and are perceived by voters to be competent in these areas (Green and Jennings 2012), governing parties are less likely to change their priorities right before an election. Greene (N.d) furthermore distinguishes between governing and opposition parties likelihood of expanding the issues 8

H3: Government parties will be less responsive to voters than opposition parties. Turning to a final distinction among parties, niche parties, defined by Meguid (2005, 347 8) are parties which reject the traditional class based orientation of politics, whose policies do not coincide with existing lines of political division, and which limit their issue appeals. 5 Niche parties are moreover classic policy seekers, which value their policy goals over any office considerations. They often formed because existing parties were not responsive to the issue priorities of part of the electorate (Kitschelt 1988), as discussed above regarding the formation of the green parties, and thus are focused more on the very policies that lead to their formation. Unlike mainstream parties, niche parties may make decisions, such as choosing to stay out of government, because of policy differences or feeling constrained by the policy considerations of the dominant party, which run counter to the traditional expectation of parties as office seekers (Downs 1957; Riker 1982). These parties are largely unresponsive to the issue agendas of other parties (Meguid 2005, 2008; Adams et al. 2006; Ezrow 2008). Additionally, niche parties may be considered the archetypal ʺassociative issue ownersʺ (Walgrave et al. 2012), in that there is a spontaneous identification between issues relating to the environment or immigration, for example, and green or far right parties, respectively, in the minds of most people (see Spoon et al. N.d.). For these reasons, we argue that niche parties may be more responsive to voters on the specific issues on which they focus. 6 Our final hypothesis reflects both niche parties differences in relation to mainstream parties as well as their ownership of specific issues. H4: Across issue areas, niche parties will be no more responsive than other parties; however they will be more responsive to voters on their own issues. on which they focus in their manifestos. He finds that opposition parties increase their issue diversity as a way of expanding their appeal to voters. 5 It is important to note that there are other definitions of niche parties in the literature. Wagner (2012), for example, has proposed that niche parties are best defined as parties that compete primarily on a small number of non economic issues. 6 It is important to note that party size and niche party status are not necessarily the same. Niche parties may be small in size (most green parties) or a major political actor (such as the Freiheitliche Partei Österreichs). Small parties, moreover, may be niche parties, but may also be mainstream parties (such as the Swedish Kristdemokraterna). See Spoon (2011) for a further discussion. 9

3. Research Design In this section, we explain in detail how we constructed the dataset to empirically test our hypotheses. We first discuss the measurement of the dependent variable party issue responsiveness before we explain the selection of the data and the operationalization of the independent variables. 3.1. Assessing party issue responsiveness across countries To measure the responsiveness of political parties to the issue priorities of voters, we examine the congruence of voter and party attention on different policy issues. More precisely, we empirically compare the attention that voters pay to different policy issues at t 1 with the policy issues that parties emphasize at t0. In order to measure issue responsiveness of political parties, we thus include a time lag between the issue priorities of voters and parties assuming that political parties need some time to process changes in issue salience to voters and to adjust their issue priorities to these issue attention changes. 7 We measure the issue attention of political parties by the emphasis that they place on different policy issues in their election manifestos. Manifestos are an ideal source for mapping the issue priorities of political parties. Parties regularly draft manifestos for national elections and clearly lay out their stance on different policy issues to signal their policy commitments to voters. They constitute a rich data source that has been used by a wide variety of scholars to measure the salience of policy issues for parties (see also Budge and Laver 1993; Bäck et al. 2011, Klüver and Spoon N.d.). We rely on information extracted from election manifestos by the Comparative Manifesto Project (CMP) (Klingemann et al. 2006; Volkens et al. 2012). The CMP has generated the most comprehensive and most widely used dataset on parties policy priorities by applying a manual content analysis to election manifestos. Human coders have divided election manifestos into units of analysis (so called quasi sentences ) and have allocated these quasisentences to policy categories specified a priori in a coding scheme. The attention that political 7 We have also repeated the analysis using data on voter issue attention at t0 and the results are largely the same. 10

parties pay to policy issues is measured by taking the percentage of quasi sentences devoted to a certain issue area (e.g. the percentage of quasi sentences of a manifesto devoted to environmental protection). In order to measure the attention that voters pay to different policy issues we rely on the socalled most important problem/issue (MIP) question included in a number of different national and cross national surveys. 8 This question asks respondents to indicate the most important problem/issue that their country is currently facing in their personal opinion. This question is a standard tool for measuring issue priorities of voters that has been used by a wide variety of scholars in comparative politics (e.g. Hobolt and Klemmensen 2008; Jennings and John 2009; Chaqués Bonafont and Palau 2011). In order to assess party responsiveness, we measure the attention that parties pay to different policy issues at t0 (national election at to) while measuring the issue priorities of voters at t 1 (previous national election). As there is a long standing debate in comparative party research whether parties respond to the entire electorate or only to their supporters or to specific groups (e.g. Dalton 1985; Miller and Schofield 2003; Adams et al. 2006; Hobolt and Klemmensen 2008; Ezrow et al. 2011), we generate two different variables that capture the issue attention of voters. First, we estimate issue attention of all voters by taking the percentage of all survey respondents in a given national election study that indicated that a certain policy issue is most important for them as a measure for the attention of voters to this policy issue. Second, we measure issue attention of supporters of a political party by taking the percentage of respondents who have voted for this party that consider a given policy issue as most important. We combine the most important problem/issue 8 We acknowledge that most important issue and most important problem are not necessarily the same question (see Wlezien 2005 and Jennings and Wlezien 2011 for a more recent discussion of how there is little difference between the two questions) and that MIP questions do not always accurately the public s preferences (see Jennings and Wlezien 2012). However, as these are the best cross national questions available to gauge policy priorities of citizens, we use both most important issue and most important problem questions in order to have as much of a cross national time series as possible (see Jennings and John 2009, 844). 11

question with a survey question asking respondents about their vote choice in the last national election to identify party supporters. To test whether political parties respond to the issue priorities of voters, we matched the policy categories included in the CMP database with the responses to the MIP question included in national election studies. We grouped the CMP categories and the public opinion data into 13 common issue areas which allows us to directly assess whether parties responded to the changes in voter attention to a common set of issue areas. Table 1 shows the CMP policy categories that we included in each issue area. TABLE 1 ABOUT HERE In order to empirically test our hypotheses, we analyze party responsiveness in 18 European democracies. 9 We initially started with the 27 member states of the European Union and checked the availability of national election studies including the most important problem question since the 1970s. For some countries, there was no election study conducted during our time period of study (e.g. Cyprus, Greece, Lithuania, Romania), while for others, the question was significantly different from the typical open ended most important problem/issue question and thus answers would not be comparable (e.g. Belgium) or the only available election study that included the MIP question was the latest election available in the CMP data (e.g. Estonia). For some of these 18 countries, however, an MIP question was not asked in every parliamentary election survey (for example, Czech Republic, France, Hungary, Ireland, Italy, Slovenia, and Spain). We therefore only include the elections for which we have a comparable MIP question from the previous election survey. If the MIP question was not asked in a previous survey or if there was no previous election study, then we drop this election from the dataset. Our data 9 The countries included in our analysis are Austria, Bulgaria, the Czech Republic, Denmark, Finland, France, Germany, Hungary, Ireland, Italy, the Netherlands, Poland, Portugal, Slovakia, Slovenia, Spain, Sweden, and the United Kingdom. 12

comes from the Comparative Study of Electoral Systems and from various additional national election studies. 10 See the Appendix for a list of election studies used for the most important issue/problem data. In all, we examine the issue responsiveness of 130 parties across 13 issue areas in 63 national elections from 1972 to 2011. 3.2. Measuring independent variables To understand how party characteristics condition responsiveness, we include three variables in our models. First, we include a variable for party size, which is measured as the vote share of the party in the previous national election. Next, we include a variable for governing party status. A party is considered a governing party if it was in government at the time of the election under consideration. Third, we code parties as niche or mainstream parties. We use Meguid s (2005) definition of niche party and code green, regional, and far right parties as such. The data on party size was obtained from the CMP dataset and the government participation data comes from the ParlGov dataset (Döring and Manow 2012). We also include several control variables. First, at the party level, we have a variable that measures the party s vote share differential from the last election to the current election, measured as the difference between the party s vote share at time t 1 and time t0. We argue that a party that has gained votes may be less responsive than one that has lost votes. We measure the variable in this way as a party will compare its status in current polls to its vote share in the previous election to determine the issues on which it will focus. Although the election has not happened yet, we believe the election result can be used as a proxy for how the party is doing in the polls. It makes less sense to argue that a party is determining the issues on which to focus in the current election based on the difference in its vote share in the past two elections as polling data is widely available and parties therefore have a good sense of their current voter support. 10 We only use MIP data from national election studies and do not rely on polling data to generate a comparable dataset across countries. All the national election studies we include in our dataset rely on open ended questions. 13

Next, we include five party system control variables. All of these variables condition how responsive a party will be to voter demands. First, we include a variable that measures average issue salience, for the party system in a given election. We posit that parties issue emphasis is not only influenced by voters issue priorities, but also by those of other parties. If other parties are focusing on an issue, then we would expect all parties in the system to do the same. Similar to the voter issue attention variable, we used the party system s average issue attention from the previous election. Second, we control for average district magnitude. We argue that the larger the magnitude, the less responsive parties will be as they do not need to appeal to as many voters to be elected. We draw these data from Golder (2006) and add the data after 2000. Finally, we control for voter turnout. We argue that a high turnout election can function as a proxy for the salience of the election. When an election is more salient, parties are more likely to respond. For this reason, we use turnout from the election under consideration instead of turnout in the previous election, which would not be a proxy for salience in the current election. 11 These data come from the International IDEA database. We also control for whether a country has compulsory voting as this may affect how a party uses voter turnout as a proxy for the importance of the election. See the Appendix for descriptive statistics for all of the variables included in our models. 4. Data Analysis In this section, we empirically test our theoretical expectations on the basis of the newly constructed dataset on issue responsiveness of political parties. After briefly discussing the specification of the statistical model, we present multivariate regression analyses testing our hypotheses step by step. 11 Re running our analyses with the lagged voter turnout variable yields similar results. As many of the countries which used compulsory voting are not included in our dataset because of data availability, the dichotomous variable for compulsory voting only refers to Italy before 1993. 14

To assess the impact of the explanatory variables on issue responsiveness of political parties, the special structure of the data has to be taken into account. Our dataset comprises information about the responsiveness of 130 parties in 18 European countries across 13 issue areas from 1972 until 2011. The observations in our dataset are therefore not completely independent as assumed by ordinary least squares (OLS) regression. In order to acknowledge the hierarchical structure of the data, we estimate OLS regression models with clustered robust standard errors by creating a cluster variable that takes into account that our observations are simultaneously nested into parties on the one hand and into issue areas on the other hand. In addition, one might also argue that the results vary across countries. We have thus also estimated an OLS model that includes country fixed effects. However, the results are substantially the same and none of the country dummies has a statistically significant effect. We therefore only present the most parsimonious model accounting for the clustering into parties and policy issue areas. Our dataset is furthermore characterized by a time component as we analyze succeeding elections in the different countries in our sample. As a result, the observations in our dataset are not only clustered into parties and issue areas, but they might also be dependent over time. In order to control for potential autocorrelation induced by the time series structure of the data, we include the lagged dependent variable (Beck & Katz 1995, 1996; Beck 2001). Table 2 presents the results of the multivariate OLS regression model examining the responsiveness of political parties to the entire electorate. Model 1 only includes the main effects for our explanatory and control variables. Model 2 then adds interaction effects for the three party type hypotheses relating to party size, government status and niche party and voter issue attention at t 1. Finally, model 3 includes interaction effects for all explanatory and control variables. In line with our first hypothesis, model 1 indicates that political parties are indeed responsive to the issue priorities of the electorate as voter issue attention at t 1 has a statistically significant effect on party issue attention at t0. Hence, political parties adjust their issue priorities in response to the policy concerns of voters. This finding confirms previous work that parties are not only responsive with regards to the policy positions they adopt on the left/right 15

dimension (e.g. Adams et al. 2004, 2006), but also listen to the voters with regard to the issues they prioritize (Spoon and Klüver N.d.). The larger the attention that voters pay to a policy issue at t 1, the larger the emphasis that parties pay to this issue at t0. More specifically, if voter attention to a policy issue at t 1 increases by one percentage point, the attention that parties pay to this issue at t0 on average increases by 0.046 percentage points. TABLE 2 ABOUT HERE However, we have theorized that issue responsiveness is not equal, but that it is conditioned by party characteristics. We have thus included interaction effects between voter issue attention at t 1 and party size, government and niche party status in model 2. In line with hypothesis 2, the empirical analysis shows that party size is positively related to issue responsiveness as the interaction effect exhibits a statistically significant positive relation. Hence, large parties, which are typically mainstream parties, are more responsive to voters than smaller parties. Similarly, the OLS regression analysis also provides empirical support for hypothesis 3: the interaction effect between voter issue attention at t 1 and government status is statistically significant and negative. Thus, government parties are on average less responsive to voters than opposition parties. Since both party size and government status exhibit statistically significant effects, it is important to note that even though large parties are generally more responsive to voters than small parties, large parties are less responsive to voters if they are in government at the time of the election. Finally, we find empirical support for the first part of hypothesis 4, that niche parties will be no more responsive than other parties. As expected, the coefficients for the niche party variable in models 1 and 2 are not significant. The empirical findings for party size, government and niche status are robust across model specifications as the effects hold no matter when solely including interaction effects for these explanatory variables (model 2) or all independent variables in the model (model 3). As we have discussed earlier, there has been a long standing debate about the question of whether parties respond to the entire electorate or just to their own supporters or a narrower 16

group of voters and whether this varies across party type (e.g. Dalton 1985; Miller and Schofield 2003; Adams et al. 2006; Hobolt and Klemmensen 2008; Ezrow et al. 2011). We have thus repeated the OLS regression analyses by including only the attention that party supporters paid to policy issues at t 1 instead of looking at the attention of all voters to policy issues at t 1. Table 3 displays the results of the analysis including only the main effects (model 4), including the main effects and additional interaction effects for party supporter issue attention at t 1 and party size, government and niche status (model 5) and for interactions with all independent variables (model 6). TABLE 3 ABOUT HERE The analysis examining responsiveness to party supporter issue attention confirms the findings of the analysis based on the entire electorate. In line with our first hypothesis, the analysis indicates that political parties respond to the issue priorities of party supporters. Similarly, the findings relating to party size, government and niche party status are equally confirmed by the analysis of party responsiveness to their supporters. Large parties are generally more responsive to their supporters than small parties and government parties are on average less responsive to the policy priorities of their supporters than opposition parties, while niche party status does not exhibit a statistically significant effect. Hence, the empirical analysis indicates that party responsiveness is similar to the entire electorate and to party supporters. This may partly be due to similar issue priorities among voters. The attention that all voters pay to policy issues correlates at 0.87 with the policy priorities of party supporters. This finding demonstrates the focus on office that parties have as parties generally respond to shifts in issue salience to citizens across all policy issues to attract voters (Downs 1957; Strøm and Müller 1999). Larger parties not only seek the support of their own supporters, but also that of the entire electorate; whereas, governing parties are no more responsive to their own voters than to the electorate as a whole when in government, as they are constrained by the nature of office. The result also confirms that niche parties are no more likely to listen to their own supporters issue priorities in general than other parties. 17

In order to further illustrate how party responsiveness varies with party size and government status, we have computed marginal effect plots as recommended by Brambor et al. (2006).Figure 1 demonstrates the issue responsiveness of political parties to all voters (figure 1a) and to their supporters (figure 1b). More precisely, the figure shows the marginal effect of voter issue attention at t 1 on party issue attention at t0 as the size of political parties varies. The solid lines indicate the point estimates of the marginal effect while the dashed lines indicate the 95 per cent confidence interval. The figure clearly demonstrates that large parties are indeed more responsive to issue priorities of (a) all voters and (b) their supporters across all levels of party size. FIGURE 1 ABOUT HERE Figure 2 plots how government status affects issue responsiveness of political parties. It presents the marginal effect of voter issue attention at t 1 on party issue attention at t0 for opposition and government parties along with a 95 per cent confidence interval. Figure 2a is estimated on the basis of the attention of all voters while figure 2b is estimated on the basis of the policy priorities of party supporters. In line with our theoretical expectations, the figure demonstrates that government parties are less responsive to (a) all voters and (b) their supporters than opposition parties. FIGURE 2 ABOUT HERE In the second part of hypothesis 4, we posited that niche parties will be more responsive to voters in the issue areas that they own. Since niche parties compete primarily in policy areas that are important to them, such as green parties in the environmental issue domain (Meguid 2005, 2008; Spoon 2012), they should largely be indifferent to policy concerns of citizens in other issue areas. In order to empirically assess whether niche parties are indeed more responsive to voters in the policy areas they own, we have estimated two additional OLS regression models in which we test whether green parties are more responsive to voters in the environmental policy 18

domain. We have chosen to test this hypothesis with regard to green parties and environmental policy as this issue ownership relationship is theoretically and empirically the most distinct example of a policy area owned by a specific party type (Meguid 2005; Spoon et al. 2012) and it furthermore can be considered an associated issue (Walgrave et al. 2012). In addition environmental policy is a typical valence issue (Stokes 1963). When a voter responds that the environment or pollution is the most important issue, it is clear what the position of the voter is a voter would not state that the environment is the most important issue in the election if she was opposed to environmental protection or reducing air pollution, for example. We can then easily match the voters issue priority with the CMP codebook which includes two policy categories that are both supportive of environmental policy and conservation (per 416 and per 501). Thus, there is a clear match between the voters issue priority and the content of the manifesto as measured by the CMP. Because of the nature of other issue areas that voters offer as the most important issue in the national election surveys, it is not clear which position voters have on these issues and thus which CMP categories and parties to match it with. For example, when voters respond that Europe, European integration, or European policy is most important, we cannot know if they are Europhile or Euroskeptic and thus cannot use this as a proxy of the Euroskeptic issue, which could be considered as the owned issue of Euroskeptic parties. Moreover, many large mainstream parties may also be considered Euroskeptic, but are not the perceived owners of the issue. Thus, for all of these conceptual and methodological reasons, we have chosen to look at the responsiveness of green parties to voters on the environment issue. TABLE 3 ABOUT HERE Table 3 presents the results of the empirical analysis. Instead of including a niche party dummy, we have controlled for green parties more specifically in the model. In addition, we have included a dichotomous variable indicating issue responsiveness in the environmental issue area as well as two way interactions between voter issue attention at t 1 and green parties and environmental policy and three way interaction between voter issue attention at t 1, green parties and environmental policy. Model 7 is estimated on the basis of issue attention of the 19

entire electorate and model 8 is based on the issue attention of party supporters. In both models 7 and 8, we see that the interaction between the green party dummy variable and both voter and supporter issue attention on all issues other than the environment, is not significant. This confirms our expectation that green parties, as niche parties, are not any more responsive than other parties are on a non owned issue. However, when we look at how green parties respond to their supporters attention to the environmental issue, which is an owned issue, we see a positive and significant coefficient. The larger the attention that supporters of green parties pay to environmental protection at t 1, the larger the emphasis green parties pay to environmental policy at t0. Hence, niche parties have clear issue and voter priorities. They do not adjust their policy priorities with regard to all voters, but only with regard to their specific supporters. Our findings follow those of Ezrow et al. (2011), who showed that niche parties respond to shifts in the mean position of their own voters rather than the entire electorate. What is more, green parties care in particular about the issue they own and they are more responsiveness to the policy concerns of their supporters in this particular policy domain than in other policy areas. This finding thus confirms the second part of hypothesis 4. FIGURE 3 ABOUT HERE Figure 3 presents simulated predicted values for the responsiveness of political parties in the environmental policy domain to illustrate this issue ownership relationship (King et al. 2000). The gray lines indicate the responsiveness of green parties to party supporter concerns regarding environmental protection while the black lines indicate the responsiveness of other parties. The solid lines indicate the point estimates while the dashed lines indicate the 90 per cent confidence interval. Figure 3 clearly shows that green parties are considerably more responsive to the attention that party supporters pay to environmental protection than other parties. While the attention that green parties pay to environmental protection at t0 clearly increases with the attention that their supporters pay to environmental policy at t 1, the issue attention of other parties remains largely unaffected by their supporters concerns in this policy area. When lagged party supporter issue attention is at its mean (where 5.31% of party 20

supporters responded that the environment was the most important issue), green parties issue attention on the environment issue is 8.94 compared to 5.86 for other parties 12, holding all other variables at their mean (continuous variables) and median (categorical variables) values. The confidence intervals for the green parties responsiveness is fairly large given the small number of observations in our dataset (N=312). However, despite the large confidence intervals of green party responsiveness, they hardly overlap with the responsiveness of other parties which further corroborates the important issue ownership relationship between green parties and environmental policy concerns of their supporters. Finally, with regard to the control variables included in the empirical analysis, the OLS regression models presented in table 2 and in table 3 only indicate the following statistically significant relationships. First, parties respond not only to voters, but also to the policy priorities of other political parties. The larger the average attention that parties pay to policy issues at t 1, the larger the attention that parties pay to these policy issues at t0 (see model 1 and model 4). However, at the same time, the OLS regressions indicate a negative interaction effect between average party issue salience and voter attention. Hence, the larger the average attention that parties and voters paid to a policy issue at t 1, the smaller the responsiveness of parties to that policy issue at t0 (see model 3 and model 6). 13 Second, the analysis demonstrates a systematic effect of average district magnitude. The larger the average district magnitude, the lower the responsiveness of parties to voters (see model 3 and model 6). This follows our expectation that with larger district magnitudes, parties can be less responsive to voters issue priorities and chose to focus on issues about which they prioritize, and still gain representation. 12 These numbers represent the percentage of the parties manifesto that focuses on the environment issue area. 13 This counterintuitive finding may be because of the lag in the data. Parties may not be responsive to other parties and voters together based on their issue priorities from the past election, but are responsive to parties priorities in the current election. Re running the analysis with average salience at t0, the interaction effect is positive, but not significant. 21

4. Conclusion How do party characteristics influence responsiveness to voters issue priorities? Building on previous research, which has demonstrated that parties respond to shifts in voters ideological positions (e.g. Adams et al. 2004, 2006, 2009; Ezrow and Hellwig 2011) and that governments respond to the issue priorities of voters (e.g. Stimson et al. 1995, Jones and Baumgartner 2004; Hobolt and Klemmensen 2008; Chaqués Bonafont and Palau 2011), we find that parties responsiveness to voters issue priorities is conditional on several characteristics, namely party size and party type. We show that although larger parties are more responsive to the issues voters prioritize, governing parties are less responsive, thus highlighting the constraints that parties are under when they participate in government. We also demonstrate that niche parties are no more responsive than other parties in general; however, when we examine niche party responsiveness on their own issues, we find that that they are more responsive to their supporters. Interestingly, looking at non niche parties, we find no difference in party responsiveness to the electorate in general as compared to party supporters. Thus, large parties are still more responsive compared to small parties even when looking only at party supporters. Niche parties, however, are more responsive to their own voters on their own issues, thus further highlighting their distinctiveness (Adams et al. 2006; Meguid 2005, 2008; Ezrow et al. 2011; Spoon 2011). Our findings have important implications for understanding political representation and political competition. First, if larger parties are more responsive than smaller parties, voters will continually be more likely to vote for them, thus perpetuating their larger party status and reducing smaller party representation. In addition, as party size often determines the resources that parties receive (Katz and Mair 1995), smaller parties are more likely to remain small. Second, although we might expect smaller parties to be more responsive to their own voters because they have chosen to be more ideologically focused (Kitschelt 1989; Cox 1997), this is not generally the case. We do, however, find niche parties to be more responsive to their own voters 22