Partisanship and Satisfaction with Democracy:

Transcription

1 Partisanship and Satisfaction with Democracy: A Longitudinal Analysis of the Linkage between Citizens and the State Anna Kern, PhD student Centre for Citizenship and Democracy, KU Leuven, Belgium anna.kern@soc.kuleuven.be Paper prepared for the ECPR General Conference University of Glasgow 3-6 September 2014 Abstract There are a number of studies indicating that partisan ties are weakening. However, we know less about what this means for citizens attitudes towards the party system and democratic procedures more generally. Relying on data from the German Socio-Economic Panel, we conduct a longitudinal multilevel analysis to explore whether the effect of party identity on satisfaction with democracy has changed over a period of 23 years ( ). Our results show that while partisanship and satisfaction with the performance of the regime declined in this period, the effect of party identity on satisfaction with the functioning of democracy has strengthened. Thereby, it does not matter whether citizens support the party that won or lost the previous elections. We conclude that by providing mechanisms of identification, political parties still play an important role in linking citizens with the state. 1

2 1. Introduction Party identification 1 has proven to be one of the most valuable elements in understanding the political behavior of contemporary publics (Dalton, 2014, p. 134). Partisanship implies that citizens develop a long-term psychological association with a political party (Campbell, Converse, Miller, & Strokes, 1960; Green & Palmquist, 1990). These emotional partisan ties have shown to be essential in various ways: First of all, party identification is highly functional for the individual citizens as it helps them to orient themselves in the political world (Dalton, 2004; Dassonneville & Grieb, 2014). Party attachments render it easier for citizens to process political information, to judge and evaluate political occurrences and to determine whether they are in favor or against certain issues and policies. Furthermore, party attachment affects political participation patterns and serves as powerful driving force for voting behavior (Bartels, 2000, 2002; Lewis-Beck, Jacoby, Norpoth, & Weisberg, 2008; Miller, 1991). Second, in addition to the functional value of partisanship for the individual citizen, party identification is claimed to structure political behavior and political attitudes more broadly, providing an attitudinal linkage between citizens and the state (Hooghe & Kern, 2013). It is argued that partisan ties foster support not only for a particular political party, a specific issue, policy or candidate but also more generally for the party system (Clarke & Stewart, 1998; Dennis & Owen, 2001), political values (Goren, 2005) and democratic procedures (Dalton & Wattenberg, 2002; Dalton, 2004). As Goren, (2005, p.882) puts it: [ ] party identification shapes a number of abstract beliefs about the good and just society. In this study we focus on this second function of partisanship, exploring the link between party identification and support for the functioning of existing democratic procedures. As considerable evidence indicates that partisan attachments have been weakening (Dalton, 2007; 1 Following common practice, we use the terms partisanship, partisan identity, partisan ties, party identification, and party attachment interchangeably. 2

3 Dassonneville, Hooghe, & Vanhoutte, 2012), we investigate whether the link between party identification and satisfaction with the way democracy works has changed over time. Thereby this study seeks to contribute to the understanding of the role that parties play in linking citizens with the state. We claim that that along with the weakening of partisan ties also the link between those ties and satisfaction with the performance of the political regime has been declining over the years. To test this hypothesis we rely on longitudinal data from the German Socio-Economic Panel (SOEP), which allows us to study the development of the effect of partisanship on satisfaction with democracy over a period of 23 years ( ). Although we assume that our hypothesis is true for all advanced representative democracies, we argue that Germany represents a good case to investigate the development of the impact of partisanship on support for the functioning of democratic procedures. The German Grundgesetz grants political parties a privileged role in the political process and party identification has been important for the consolidation of German democracy since 1949 (Rudzio, 2011). However, also in Germany not only the amount of partyidentifiers has steadily been decreasing (Arzheimer, 2006; Schmitt-Beck & Weick, 2001), also the strength of party identification has been weakening (Dalton, 2014). The paper is structured as follows. First, we will provide an overview of the literature on partisanship and its development over time. Second, we introduce the German Socio-Economic Panel and the methods we use, before we present the results of the analyses. The study closes with a discussion and conclusion. 2. Party identification as attitudinal linkage between citizens and the state Although there is vast amount of literature indicating that political parties are in crisis, progressively losing their core functions in representative democracy (Dalton & Wattenberg, 2002; Katz & Mair, 1992; van Biezen & Poguntke, 2014; Webb, Farrell, & Holliday, 2002; Whiteley, 2011), they still play a major role in socializing citizens into politics, providing them 3

4 with political orientation and a sense of identification (Dennis & Owen, 2001; van Biezen, 2004). Political parties have a cue-giving function and citizens rely on those party cues in order to relate to the political systems. Hence, party identification plays a major role in relating citizens with the state (Dalton & Wattenberg, 2002). The concept of partisanship, which was introduced more than 50 years ago in The American Voter (Campbell et al., 1960), proposes that citizens develop a long-standing psychological attachment to a political party. Therefore partisanship essentially describes the relationship of an individual with a group. As Campbell et al. (1960, p.121) put it, party identification characterizes the individual s affective orientation to an important group-object in his environment. Political parties thus represent groups towards which the individual potentially develops an identification. Accordingly, being entirely a matter of self-definition (Green & Palmquist, 1990, p. 874), Greene (2004) underlines that partisanship represents a self-perceived membership in a group, rather than an attitude towards a political party. This assessment is important because it allows, to relate partisanship to social identity theory (Tajfel & Turner, 1986; Tajfel, 1978). At the center of this large body of literature lies the assumption that individuals classify the world in us and them (Greene, 1999). This classification and self-perceived group membership lead individuals to internalize the norms and values that prevail in this group and subsequently, those norms and values shape their own attitudes and behaviors. Many scholars have already applied this process of social identity formation to the development of partisan ties (Gerber, Huber, & Washington, 2010; Greene, 1999; Kelly, 1988, 1989; Weisberg & Greene, 2003). Following these authors, social identity formation forms the basis of partisanship, explaining why party supporters adopt the norms and values promoted by the party. As the large majority of political parties promote not only support for their positions and candidates but also for the party system and democratic procedures more generally, party 4

5 supporters are more inclined to support the performance of the political regime (Dennis & Owen, 2001). Although, in Europe the percentage of citizens identifying with a political party seems to be stable in recent years (Hooghe & Kern, 2013), there is ample empirical evidence pointing towards an erosion of party identification in last quarter of the 20 th century (Arzheimer, 2006; Clarke & Stewart, 1998; Dalton, 2002; Dassonneville & Grieb, 2014; Dassonneville et al., 2012). What is more, the weakening of partisan ties becomes apparent not only in the declining percentage of citizens who indicate that they have some kind of party identification, the erosion also becomes visible regarding the strength of citizens party affiliations. It seems that partisanship becomes, not only more scarce but also weaker (Arzheimer & Schoen, 2005, p own translation). Nevertheless, different explanations have been advanced to explain why party identification is eroding. Some authors claim that partisanship is declining because citizens gained more political skills while at the same time, mass media became widely available which decreased information costs severely (Albright, 2009; Dalton & Wattenberg, 2002; Dalton, 2007). As a consequence citizens became progressively more capable to orient themselves in the complex political landscape without relying on partisan ties. According to this cognitive mobilization theory, the rise of education levels and the spread of mass media made party attachment increasingly obsolete. Following this theory, a direct link between rising education levels and spreading mass media on the one hand, and party identification on the other hand is made. Other authors, however, claim that the process of cognitive mobilization did not directly affect partisan attachments. These scholars claim that partisanship has been eroding because, rising education levels led to changes in the social stratification of society which in turn weakened partisan ties. The argument is that with rising levels of education, the privileged relationship between traditional groups of citizens (for example workers or church-goers) and their corresponding parties are weakening (Bardi, Bartolini, & Trechsel, 2014). This process led to a 5

6 the blurring of class distinctions and to the emergence of the post-materialist cleavage (Bardi et al., 2014) 2. As citizens identify less with these traditional groups and as new value systems emerge, partisanship is waning (van Biezen & Poguntke, 2014). This argument is confirmed by Arzheimer (2006) who shows that the decline of partisanship in Germany is mainly caused by the weakening of traditional social ties. In the first two decades of the Federal German Republic its voters predominantly identified with the two major parties: While church-goers identified mostly with the Christian Democrats (CDU/CSU), blue-collar workers represented the typical electorate for the Social Democrats (SPD). Party identification thus evolved predominantly along class and religion, the two most important cleavages in German society during that period (Pappi, 1973). However, with the weakening of these cleavages partisanship started to decline from the 1970s onwards (Arzheimer & Schoen, 2005). We claim that along with declining party identification also the link between party identification and satisfaction with democracy has weakened. Studying the development of satisfaction with democracy in Germany is important for two reasons. First of all, representing an indicator of regime performance (Linde & Ekman, 2003), satisfaction with democracy is closely related with the legitimacy of the democratic system (Blais & Gélineau, 2007). This link is twofold (Scharpf, 1970, 1999): When citizens perceive the decision-making processes as open, transparent, participatory and inclusive (government by the people) the regime s input-legitimacy is strengthened. By contrast, its output-legitimacy increases with the effectiveness of the regime s decision-making process and the provision of goods and services (government for the people). Satisfaction with the functioning of democracy relates thus arguably to both forms of legitimacy. The second reason, why it seems important to investigate the development of satisfaction with democracy is that it appears to have a functional value for the working of the democratic system. 2 This explanation also seems to connect most directly to the original conceptualization of partisanship, as it was presented in the American Voter (Campbell et al., 1960). As Greene (2004) points out, already Campbell and his 6

7 As Blais and Gélineau (2007) point out, satisfaction with democracy ties in with what Dalton (2004) describes as the evaluative dimension of political support. Dalton furthermore expects high levels of this support to go along with voluntary compliance with regime directives, regulations and procedures. Accordingly, eroding levels of satisfaction with regime performance jeopardize the effective functioning of the democratic system, as political transaction costs rise. As Dalton (2004, p. 159) indicates democracies function with a minimal coercive force because of the legitimacy of the system and the voluntary compliance of the public. Hence, high levels of satisfaction with democracy seem essential for the functioning of democracy and as Dennis and Owen (2001) show, citizens with strong party attachments, evaluate the functioning of the system more positively. However, in this article, we assume that along with the decline of partisan ties, the relationship between party affiliation and satisfaction with democracy has weakened. This assumption leads to the following hypothesis: Hypothesis: The link between party identification and satisfaction with democracy has declined over the years. This hypothesis will be tested using data from the German Socio-Economic Panel. We will briefly describe these data in the following paragraph, before turning to measurements and method. 3. Data, Measurements and Method 3.1. Data To test the hypothesis we rely on data from the German Socio Economic Panel (SOEP), which is a representative, longitudinal study of German households. Started in 1984, the SOEP collects annual information on households and on household members, covering a period of thirty years (Wagner, Frick, & Schupp, 2007). In each of these years almost 11,000 households were sampled and all adult members of the household (every member that is at least 17 years old) colleagues underlined the importance of membership in social, religious or ethnic groups for the identification with a 7

8 is interviewed (Wagner et al., 2007). The question targeting respondents satisfaction with democracy is asked five times in 1987, 1988, 1989, 2005 and However, due to panel attrition and panel refreshments not every respondent participated in all five years. In order to still cover this time period of 23 years appropriately, we only included respondents in the analysis that have taken part at least in one wave in the 1980 and in in one of the two more recent waves (either in 2005 or in 2010), leaving us with a sample of 4,198 respondents and 18,940 respective observations 3. It has to be noted that by applying this restriction of having one observation from the 1980s and one from 2005 or 2010, we can only take respondents into account that live in former West Germany, because respondents living in former East-Germany are only in the sample since the reunification in However, excluding those respondents is commonly done, because political cultures in former East and West Germany still differ substantially (Arzheimer, 2006) and as Dalton (2012) shows also the strength of partisan ties differs considerably in the two geographical areas Measurements and Method The wording of the question from which we derive the dependent variable, satisfaction with democracy, has changed during the 23 years in which it has been asked. In 1987, 1988 and 1989, the SOEP questionnaire contained the following question: What about the democracy in the Federal Republic of Germany? How satisfied are you all in all with the way the democracy exists in the Federal Republic of Germany? Response categories are: 1) very satisfied, 2) moderately satisfied, 3) somewhat satisfied, 4) somewhat dissatisfied, 5) moderately dissatisfied, 6) very dissatisfied or 7) don t know 5. In 2005 and 2010 the question read as follows: How satisfied are you with democracy as it exists in Germany? Here respondents answered by attributing a score political party. 3 About 1 percent of those respondents have been observed twice, 5 percent was observed three times, about 35 percent four times, and about 59 percent of those respondents took part in all five waves. 4 In the analysis we include only data originating from respondents aged 17 and older who live in realized households (netto 10-19) For more information, see SOEP-Documentation PPFAD ( 5 Respondent answering 7) Don t know, were not considered in the analysis (13.5% of the observations in the sample). 8

9 on a 11 point scale ranging from 0 indicating low satisfaction to 10 indicating high satisfaction. Because both questions tap the level of support for how the democratic regime works in practice (Linde & Ekman, 2003, p. 405) we considered them comparable and adequate to answer our research question. We harmonized them in two steps. First, we reversed the answer categories of the question asked in the 1980s so that higher values indicate a higher satisfaction with the working of democracy for both items. Second, we rescaled those six answer categories to a scale ranging from 0 to 10. The resulting variable has a mean of 5.61 (SD= 2.44). The average satisfaction with the functioning of the democracy declined steadily from 6.82 (SD=2.23) in 1987 to 5.35 (SD=2.19) in The independent variable of interest in this study is party identification. In contrast to the dependent variable the questions that allow us to determine an individual s party attachment have been asked in every year since Party identification can be determined using three questions. First, respondents were asked Many people in Germany are inclined to a particular party over time even if they vote for another party now and then. How about you: Are you inclined to a particular party in Germany? Second, those respondents who answered with Yes on the first question were asked Which party are you inclined to? The third question then aims at the strength of their party identification, reading How strong are you inclined to this party? Based on the first and the third question we created a categorical variable for party identification, that discriminates between 0) respondents that have no party attachments, respondents with 1) very weak, 2) fairly weak, 3) moderate, 4) fairly strong and 5) very strong party attachments. As it has been shown by other scholars using the same data (Arzheimer & Schoen, 2005; Dassonneville & Grieb, 2014), party identification is declining. This decline becomes visible in absolute numbers as well as in strength: While in 1987, 31.5 percent of the respondents indicated that they had no party attachments, the amount of apartisans rose to about 50 percent in 2005, before it slightly declined to about 47 percent in Also, the percentage of respondents who perceived themselves as fairly strong and very strong partisans declined from about 33 percent in 1987 to 9

10 22.5 percent in Following our argumentation we expect that also the link between party attachment and satisfaction with democracy weakened. However, we need to control for other individual characteristics might also affect the evaluation of the functioning of democracy. We control for individual background characteristics. Some of them are assumed to be timeinvariant: gender, year of birth and religion. Immigrants and non-nationals might be confronted with discrimination, which is why they might be less satisfied with the functioning of the democratic system. We therefore control for whether a respondent is born in Germany (as timeinvariant characteristic) and whether the respondent has the German nationality (as time-variant variable). Additionally we control for characteristics that can change over time: A respondent s socio-economic status is one of those variables. We measure socio-economic status with a respondent s level of education and her or his satisfaction with the household s income 6. Previous studies show that citizens with higher levels of education (Armingeon & Guthmann, 2014) and income (Anderson & Guillory, 1997) are more satisfied with the functioning of democracy. Also, Anderson and Guillory (1997) show that interest in politics is an important variable when studying satisfaction with democracy in Germany. As we expect a non-linear effect of political interest on satisfaction with democracy, assuming that the very interested citizens as well as the not at all interested citizens are less satisfied with the performance of democracy compared to the average interested citizens, we introduce also its quadratic term. Finally, we include a variable for time in the analysis. This variable is constructed by subtracting the year in which the interview was conducted from 1984 which was the first year of the survey. For the descriptive statistics of these independent variables see Table 1. The aim of this paper is to explore the whether the relationship between party identification and satisfaction with democracy has weakened over time. We thus need a method that allows us to model change over time. This longitudinal dataset contains repeated measures over time and is 6 The latter represents a proxy for actual income. However, 4.3 percent of the respondents in our sample did not indicate their household income, whereas only 1 percent did not answer the question on satisfaction with household income. We therefore decided to use this latter, admittedly more subjective measure. 10

11 therefore hierarchically structure with observations being nested within respondents (Gelman & Hill, 2007). Consequently, we fit a multilevel model for change and include an interaction between party identification and time to test our hypothesis (Singer & Willett, 2003). Table 1: Descriptive statistics N mean sd min max Missings (obs.) in % Party Identification 17, Sex (Female=1) 18, Year of Birth 18, Born in Germany (Yes=1) 18, German Citizen (Yes=1) 18, Religion Catholic 17, Protestant 17, Other Christian Confession 17, Islam 17, Other Religion 17, No religious confession (ref.) 17, Religious Practice 18, Level of Education 18, Satisfaction with Income 18, Political Interest 18, Time 18, Source: SOEP v29, own calculations. Note: The missings are calculated with respect to the sample size of 18,940 observations. We obtain this sample size when only considering respondents that have taken part at least in one wave in the 1980 and in one of the two more recent waves (either in 2005 or in 2010). 4. Results In the following paragraphs we present the results of our longitudinal multilevel analysis seeking to answer the question whether the effect of partisanship on satisfaction with democracy changed over time. We start the analysis by fitting an unconditional means model (Model I in Table 2). While there are no independent variables included in this model, it shows the estimated variance of satisfaction with democracy within respondents ( =3.953) and across respondents ( =1.829) without taking time into consideration. This allows us to calculate the intraclass correlation coefficient, which is 0.316, meaning that 31.6 percent of the variance of satisfaction with democracy is found at the second level between respondents and consequently almost 70 percent 11

12 is found within respondents. Moreover the model informs us that the grand mean of satisfaction with democracy is 6 7. In the second model, which is an unconditional growth model, we add time as a first-level predictor. The effect of time confirms what we have found when looking at the average satisfaction with democracy per year: There is a significant negative trend, indicating that satisfaction with democracy has declined during those 23 years covered by our analysis. Finally we introduce the variable of interest together with the control variables in Model III. Party identification has, as expected a significant positive effect on satisfaction with the functioning of democracy. However in order to explore whether this effect changed over time, we need to include an interaction between party identity and time. This is done in Model IV, revealing, in sharp contrast to our hypothesis, a positive effect. This means that the positive effect of partisanship on satisfaction with democracy has strengthened over the years. While partisan ties are weakening and satisfaction with the functioning of democracy declines, the effect party attachments on satisfaction with democracy became stronger. One reason for this effect might be that we did not take into account with which party respondents identify. As Anderson, Blais, Bowler, Donovan, and Listhaug (2007), Anderson and Guillory (1997) and Blais and Gélineau (2007) demonstrate, there is a satisfaction gap between the winners and the losers of the electoral competition. Voters who supported a party that won the elections tend to be significantly more satisfied with democracy than those who voted for a party that joined the opposition after the elections. On top of that, Enyedi (2014) finds that the difference between winners and losers increased in Western Europe between 1996 and So if the effect of party identification is driven by this incumbency effect and if the differences between winners and losers become more pronounced the strengthening of the effect of partisanship might be an artifact. 7 This is slightly higher than what has been shown before in the measurement section. The difference is a result of the different sample. As we attempt to keep all the models comparable, we excluded all observations from the analysis that have a missing value on one of the independent variables. 12

13 Table 2: The effect of the party identification on satisfaction with democracy over time Model I Model II Model III Model IV Intercept 5.959*** 6.652*** ** ** (0.028) (0.040) (3.689) (3.686) Time-invariant Variables (Level 2) Female (0.052) (0.052) Year of Birth 0.008*** 0.008*** (0.002) (0.002) Born in Germany (Yes=1) (0.126) (0.126) Religion ( ref. no religious confession) Catholic 0.281** 0.280** (0.101) (0.101) Protestant 0.261** 0.260** (0.098) (0.098) Other Christian Confession 0.477** 0.482** (0.177) (0.177) Islam (0.140) (0.140) Other Religion (0.532) (0.532) Time-varying Variables (Level 1) Party Identification 0.106*** 0.077*** (0.012) (0.018) Time *** *** *** (0.002) (0.002) (0.003) Level of Education 0.043*** 0.042*** (0.010) (0.010) Satisfaction with Income 0.241*** 0.241*** (0.009) (0.009) Political Interest 0.972*** 0.960*** (0.125) (0.125) Political Interest *** *** (0.025) (0.025) German Citizen (Yes=1) * * (0.148) (0.148) Religious Practice 0.137*** 0.136*** (0.025) (0.025) Time * Party Identification 0.002* (0.001) Variance Components Level 1 Within-Person ( ) (0.055) (0.048) (0.048) (0.049) Level 2 In initial status ( ) (0.070) (0.129) (0.120) (0.120) In rate of change (0.000) (0.000) (0.000) Covariance (0.006) (0.006) (0.006) Deviance 62,585 61,141 60,228 60,224 N Observations 13,992 13,992 13,992 13,992 N Respondents 3,730 3,730 3,730 3,730 Note: The dependent variable is the satisfaction with democracy. Standard errors in parentheses. Sign.: * < 0.05, ** < 0.01, *** <

14 In order to rule this possibility out we conduct an addition analysis and test first, whether party identification still affects satisfaction with democracy if we control for such a potential incumbency effect. Second, if party identification still affects satisfaction with democracy, we test whether the effect still becomes stronger over time. In Model V (Table 3) we repeat the model with party identification and the control variables (Model III in Table 2) and add thee dummy variables that denote respondents who 1) identify with the Christian Democrats (CDU/CSU), 2) the Social Democrats (SPD) and 3) a party that was in government at the moment of the interview. As Model V shows, there is indeed an incumbency effect: Citizens that identify with a party in government at the moment of the interview are more satisfied with the functioning of democracy than citizens who identify with a party that is not government or do not identify with a party at all. This result confirms previous findings that winners are indeed more satisfied with regime performance (Anderson et al., 2007; Anderson & Guillory, 1997). Furthermore, this model shows that independent of whether Christian Democrats have won or lost the elections, their supporters are more satisfied with the system than citizens who have another or no party affiliation. However, most importantly, controlling for these additional variables, general party identification has still a positive effect. Yet, what remains unclear is whether this effect still becomes stronger over time if we control for party identification with the traditional parties and identification with the incumbent. This question is explored in the following two models. Model VI shows that the effect of party identification is getting stronger over time even if we control for these additional variables. Additionally, we find in Model VII that the strengthening of the effect of partisanship also holds if we control for the evolution of the incumbency effect. Whereas the effect of party identity on satisfaction with democracy becomes stronger over the years, the strength of the incumbency effect seems to erode. While Enyedi (2014) shows that the satisfaction gap between winner and losers and its effect on satisfaction with democracy increased in Western Europe, we find the opposite result for the German citizens. The 14

15 incumbency effect on satisfaction with democracy is significant and positive but it s strength declines 8 over the years and it can therefore not explain the strengthening of the effect of general party identity. Table 3: The effect of the party identification on satisfaction with democracy over time controlling for a potential incumbency effect Model V Model VI Model VII Intercept *** (3.663) *** (3.661) (3.661) Time-invariant Variables (Level 2) Female (0.052) (0.052) (0.052) Year of Birth 0.009*** (0.002) 0.009*** (0.002) 0.009*** (0.002) Born in Germany (Yes=1) (0.125) (0.125) (0.125) Religion ( ref. no religious confession) Catholic (0.100) 0.197* (0.100) 0.201* (0.100) Protestant 0.240* (0.097) 0.239* (0.097) 0.240* (0.097) Other Christian Confession 0.424* (0.176) 0.429* (0.176) 0.433* (0.176) Islam (0.138) (0.138) (0.138) Other Religion (0.530) (0.530) (0.530) Time-variant Variables (Level 1) Party Identification 0.048* (0.019) (0.023) (0.024) Time *** (0.002) *** (0.003) *** (0.003) Level of Education 0.042*** (0.010) 0.042*** (0.010) 0.041*** (0.010) Satisfaction with Income 0.239*** (0.009) 0.239*** (0.009) 0.239*** (0.009) Political Interest 0.943*** (0.124) 0.933*** (0.124) 0.934*** (0.124) Political Interest *** (0.025) *** (0.025) *** (0.025) German Citizen (Yes=1) * (0.147) * (0.147) * (0.147) Religious Practice 0.111*** (0.025) 0.109*** (0.025) 0.107*** (0.025) Identification CDU/CSU 0.179* (0.085) 0.175* (0.085) (0.091) Identification SPD (0.072) (0.072) (0.073) Identification with Incumbent 0.460*** (0.056) 0.459*** (0.056) 0.804*** (0.110) Time * Party Identification 0.002* (0.001) 0.005*** (0.001) Time * Identification with Incumbent *** (0.006) Variance Components Level 1 Within-Person (0.048) (0.048) (0.048) Level 2 In initial status (0.116) (0.116) (0.115) In rate of change (0.000) (0.000) (0.000) Covariance (0.006) (0.006) Deviance 60,086 60,082 60,069 N Observations 13,992 13,992 13,992 N Respondents 3,730 3,730 3,730 Note: The dependent variable is the satisfaction with democracy. Standard errors in parentheses. Sign.: * < 0.05, ** < 0.01, *** < The interaction between identifying with the incumbent and time was also tested in a model in which we did not include the interaction between party identity and time (see Table 1 in the Appendix). 15

16 5. Discussion and Conclusion We started the analysis to explore whether the effect of party identification on satisfaction with democracy changed over time. Given the results of previous studies that indicated that partisan ties are weakening we argued that also the effect of partisanship on satisfaction with the performance of the democratic regime declines. However, contradicting our expectations we found that the effect of partisanship on satisfaction with democracy became stronger over the years. One possible explanation for the reinforcement of the effect could have been that the effect of partisanship is actually dependent on the kind of party with which an individual identifies. A number of studies indicated that whether one identifies with the winner or the loser of an election actually matters (Anderson et al., 2007; Blais & Gélineau, 2007; Linde & Ekman, 2003). And since this the effect of identifying with the incumbent on satisfaction with democracy is claimed to become more important in Western Europe (Enyedi, 2014), this mechanism might explain our finding. However, as we have shown in an additional test, the incumbency effect declined in strength over the years in Germany and can consequently not explain the reinforcement of the partisan effect. Therefore the main conclusion of this study is that although partisan ties are weakening, party identification remains an important factor in sustaining satisfaction with the functioning of democracy in Germany and this effect of party identification on satisfaction with democracy became even stronger during the period of observation ( ). High levels of satisfaction with democracy are important, not only because regime performance is linked to the legitimacy of the political system, but also because it lowers political interaction cost, making it easier for the government to implement legislation. Our study shows that that no matter whether citizens identify with the party that won the elections, partisans are generally more satisfied with the performance of democracy than apartisans. Consequently, parties still play an important role in German democracy, providing an attitudinal linkage between citizens and the state. 16

17 However, several limitations of this study should be acknowledged. Firstly, despite the fact, that the German SOEP seems to represent an appropriate data source for testing our research question, it only contains one item for satisfaction with democracy. Relying on this single item we can only investigate the effect of partisanship on one dimension of political support, namely regime performance. Preferably we would have had additional measures to capture political support more broadly; however the SOEP does not include any other item that could have been used for this purpose. Secondly, we have studied the relationship between self-perceived group membership and an attitude towards the functioning of democracy. Consequently, what we do not know is whether the strengthening of the effect of partisanship on satisfaction with democracy has any substantive consequence. One might for instance wonder whether the strengthening of this effect had any behavioral consequence, a question, which requires further research. Nevertheless, we believe that the study provides evidence suggesting that by providing mechanisms for identification, parties still play an important role in the contemporary democratic system. Andeweg (2003, p.151) might be right with his assessment: Religion is increasingly expressed outside churches, interest promotion is taken care of outside interest associations, such as trade unions, physical exercise outside sports clubs (Putnam s famous bowling alone ), work outside permanent employment, love outside marriage, and even gender differences are becoming divorced from sex differences. The effects of party identification however are (still) hardly imaginable without political parties. 17

18 Appendix Table 1: The effect of the identifying with the incumbent on satisfaction with democracy over time Satisfaction with democracy Model I Intercept *** (3.664) Time-invariant Variables (Level 2) Female (0.052) Year of Birth 0.009*** (0.002) Born in Germany (Yes=1) (0.125) Catholic 0.199* (0.100) Protestant 0.241* (0.097) Other Christian Confession 0.424* (0.177) Islam (0.139) Other Religion (0.530) Time-variant Variables (Level 1) Party Identification 0.047* (0.019) Time *** (0.002) Level of Education 0.042*** (0.010) Satisfaction with Income 0.239*** (0.009) Political Interest 0.949*** (0.124) Political Interest *** (0.025) German Citizen (Yes=1) * (0.147) Religious Practice 0.110*** (0.025) Identification CDU/CSU (0.089) Identification SPD (0.073) Identification with Incumbent 0.635*** (0.099) Time * Identification with Incumbent * (0.005) Variance Components Level 1 Within-Person (0.048) Level 2 In initial status (0.116) In rate of change (0.000) Covariance (0.006) Deviance 60,069 N Observations 13,992 N Respondents 3,730 Note: The dependent variable is the satisfaction with democracy. Standard errors in parentheses. Sign.: * < 0.05, ** < 0.01, *** <

19 References Albright, J. J. (2009). Does political knowledge erode party attachments?: A review of the cognitive mobilization thesis. Electoral Studies, 28(2), Anderson, C. J., Blais, A., Bowler, S., Donovan, T., & Listhaug, O. (2007). Losers Consent: Elections and Democratic Legitimacy. Oxford: Oxford University Press. Anderson, C. J., & Guillory, C. A. (1997). Political Institutions and Satisfaction with Democracy: A Cross- National Analysis of Consensus and Majoritarian Systems. The American Political Science Review, 91(1), Andeweg, R. B. (2003). Beyond representativeness? Trends in political representation. European Review, 11(02), Armingeon, K., & Guthmann, K. (2014). Democracy in Crisis? The declining support for national democracy in European countries, European Journal of Political Research. Arzheimer, K. (2006). Dead Men Walking? Party Identification in Germany, Electoral Studies, 25, Arzheimer, K., & Schoen, H. (2005). Erste Schritte auf kaum erschlossenem Terrain. Zur Stabilität der Parteiidentifikation in Deutschland. Politische Vierteljahresschrift, 46(4), Bardi, L., Bartolini, S., & Trechsel, A. (2014). Party Adaptation and Change and the Crisis of Democracy. Party Politics, 20(2), Bartels, L. M. (2000). Partisanship and Voting Behavior, American Journal of Political Science, 44(1), 35. Bartels, L. M. (2002). Beyond the Running Tally: Partisan Bias in Political Perceptions. Political Behavior, 24(2), Blais, A., & Gélineau, F. (2007). Winning, Losing and Satisfaction with Democracy. Political Studies, 55(2), Campbell, A., Converse, P. E., Miller, W. E., & Strokes, D. E. (1960). The American Voter. Chicago: University of Chicago Press. Clarke, H. D., & Stewart, M. C. (1998). The decline of parties in the minds of citizens. Annual Review of Political Science, Dalton, R. J. (2002). The decline of party identifiactions. In R. J. Dalton & M. P. Wattenberg (Eds.), Parties without Partisans: Political change in advanced industrial democracies (pp ). Oxford: Oxford University Press. Dalton, R. J. (2004). Democratic Challenges, Democratic Choices: The Erosion of Political Support in Advanced Industrial Democracies. (Oxford, Ed.). Oxford: Oxford University Press. Dalton, R. J. (2007). Partisan Mobilization, Cognitive Mobilization and the Changing American Electorate. Electoral Studies, 26( ). Dalton, R. J. (2012). Apartisans and the changing German electorate. Electoral Studies, 31,

20 Dalton, R. J. (2014). Interpreting Partisan Dealignment in Germany. German Politics, 23(1-2), Dalton, R. J., & Wattenberg, M. P. (Eds.). (2002). Parties without Partisans: Political change in advanced industrial democracies. Oxford: Oxford University Press. Dassonneville, R., & Grieb, A. (2014). How Partisanship is Strengthened in Election Times: A Longitudinal Analysis using the German Socio-Economic Panel, Paper Presented at the 13th Belgian Dutch Political Science Conference. Maastricht, June Dassonneville, R., Hooghe, M., & Vanhoutte, B. (2012). Age, Period and Cohort Effects in the Decline of Party Identification in Germany: An Analysis of a Two Decade Panel Study in Germany ( ). German Politics, 21(2), Dennis, J., & Owen, D. (2001). Popular Satisfaction with the Party System and Representative Democracy in the United States. International Political Science Review, 22(4), Enyedi, Z. (2014). The discreet charm of political parties. Party Politics, 20(2), Gelman, A., & Hill, J. (2007). Data Analysis Using Regression and Multilevel/Hierarchical Models. Cambridge: Cambridge University Press. Gerber, A. S., Huber, G. A., & Washington, E. (2010). Party Affiliation, Partisanship, and Political Beliefs: A Field Experiment. American Political Science Review, 104(04), Goren, P. (2005). Party Identification and Core Political Values. American Journal of Political Science, 49(4), Green, D. P., & Palmquist, B. (1990). Of Artifacts and Partisan Instability. American Journal of Political Science, 34(3), 872. Greene, S. (1999). Understanding Party Identification: A Social Identity Approach. Political Psychology, 20(2), Greene, S. (2004). Social Identity Theory and Party Identification. Social Science Quarterly, 85(1), Hooghe, M., & Kern, A. (2013). Party Membership and Closeness and the Development of Trust in Political Institutions: An Analysis of the European Social Survey, Party Politics, forthcomin. Katz, R. S., & Mair, P. (Eds.). (1992). Party organizations: A Data Handbook on Party Organizations in Western Democracies, London: Sage. Kelly, C. (1988). Intergroup Differentiation in a Political Context. British Journal of Social Psychology, 27(4), Kelly, C. (1989). Political Identity and Perceived Intragroup Homogeneity. British Journal of Social Psychology, 28(3), Lewis-Beck, M. S., Jacoby, W. G., Norpoth, H., & Weisberg, H. F. (2008). The American Voter Revisited. Ann Arbor: University of Michigan Press. Linde, J., & Ekman, J. (2003). Satisfaction with democracy: A note on a frequently used indicator in comparative politics. European Journal of Political Research, 42(3),

21 Miller, W. E. (1991). Party Identification, Realignment, and Party Voting: Back to the Basics. The American Political Science Review, 85(2), Pappi, F. U. (1973). Parteiensystem und Sozialstruktur in der Bundesrepublik. Politische Vierteljahresschrift, 14, Rudzio, W. (2011). Das politische System der Bundesrepublik Deutschland. Wiesbaden: VS Verlag für Sozialwissenschaften. Scharpf, F. W. (1970). Demokratietheorie zwischen Utopie und Anpassung. Konstanz: Universitätsverlag. Scharpf, F. W. (1999). Regieren in Europa: Effectiv und democratisch? Frankfurt am Main: Campus. Schmitt-Beck, R., & Weick, S. (2001). Die dauerhafte Parteiidentifikation - nur noch ein Mythos? Eine Längsschnittanalyse zur Identifikation mit politischen Parteien in West- und Ostdeutschland. Informationsdienst Soziale Indikatoren, 26(1-5). Singer, J. D., & Willett, J. B. (2003). Applied Longitudinal Data Analysis: Modeling Change and Event Occurence. Oxford: Oxford University Press. Tajfel, H. (1978). Social Categorization, Social Identity and Social Comparison. In H. Tajfel (Ed.), Differentiation Between Social Groups (pp ). London: Academic Press. Tajfel, H., & Turner, J. C. (1986). The Social Identity Theory of Intergroup Behaviour. In S. Worchel & W. G. Austin (Eds.), Psychology of Intergroup Relations (pp. 7 24). Chicago: Nelson-Hall. Van Biezen, I. (2004). Political Parties as Public Utilities. Party Politics, 10(6), Van Biezen, I., & Poguntke, T. (2014). The Decline of Membership-Based Politics. Party Politics, 20(2), Wagner, G. G., Frick, J. R., & Schupp, J. (2007). The German Socio-Economic Panel Study (SOEP) Scope, Evolution and Enhancements. Schmollers Jahrbuch, 127(1), Webb, P., Farrell, D. M., & Holliday, I. (2002). Political Parties in Advanced Industrial Democracies. Oxford: Oxford University Press. Weisberg, H. F., & Greene, S. (2003). The Political Psychology of Party Identification. In M. MacKuen & G. Rabinowitz (Eds.), Electoral Democracy (pp ). Ann Arbor: University of Michigan Press. Whiteley, P. F. (2011). Is the Party Over? The Decline of Party Activism and Membership across the Democratic World. Party Politics, 17(1),