Political Knowledge Mediator of Political Participation? Severin Bathelt July Poznań, Poland. Political Socialization and Education

Transcription

1 Political Knowledge Mediator of Political Participation? Severin Bathelt Paper presented at the IPSA 24 th World Congress of Political Science July 2016 Poznań, Poland Section: Political Socialization and Education Panel: Participation - Expectations and Challenges Abstract Political participation is fundamental for a functioning democracy. Citizens seem to fall short of the normative relevance of political involvement postulated by democratic theory. Generally, research on individual factors influencing political participation focuses either on individual resources or political socialization. I state that political knowledge can mediate effects of or is in its effects mediated by the underlying factors. Political knowledge is not seen as a simple additional individual resource, as it also constitutes a part of political socialization in the form of the manifestation of political learning and education. Research on political knowledge states that high levels of factual information always increase political participation. These findings can be contested when being controlled for levels of political trust. I use data from Allbus 2008 to review the effects of political knowledge on political participation provided that political trust is controlled for in standard models explaining participation. Keywords: Citizenship, Democracy, Knowledge, Political Participation Note: Work in Progress, please do not cite without the author s written permission! Contact Information Severin Bathelt bathelt@uni-landau.de

2 Democratic systems are designed to limit individual power and redistribute authority to the citizenry which is governed. Forms of democratic organization should desensitize against the misuse of power, include the demands of its citizens and generate decisions based on the common good. The major achievement of democratic organization is the possibility to include citizens into the decision-making process (Kaase, 1982). This requires the political participation of citizens in order to formulate demands from the political system, which can then be transferred into decisions of the governing process (Easton, 1965, p. 32). On the other hand, if these outputs enhance common good, support for the system will be generated; a very simplistic view on the reality of governing. The reality is far more different from this organizational idea. Support for democratic systems is more complex (Fuchs, Gabriel, & Völkl, 2002; Maier, 2000a), trust in democracy is decreasing (Hetherington, 1998), political participation is scarce (Berelson, Lazarsfeld, & McPhee, 1954; Campbell, Converse, Miller, & Stokes, 1960; Neuman, 1986), and political knowledge of lesser concern to the public (Converse, 1990). The idea that democracy only functions because a well-educated elite with high levels of political knowledge, acting consistently according to their attitude, has chairs in political decisionmaking might not be far from reality (Almond, 1950; Kim, 2009; Popkin, 1991; Price, 1992; Price & Zaller, 1993; Zaller, 1992). This poses a normative concern for the inclusion of the broad citizenry into the democratic process. One might understand that trust in the political system is weakened if experiences with political actions tend to be futile. But what about the implications of political trust and knowledge on political action in the first place? Are citizens less active if they think that the system does not operate on their behalf or do they tend to become more active, yet not in classical ways such as voting every couple of years? If they want to contribute, do they need to be political experts, and therefore, become part of the participatory elite? This paper sets out to explore the relation between political trust, political knowledge, and political action, as there is little research with regard to the interdependencies of these constructs. Political Partizipation Closely related to the systemic theory of democratic systems (Easton, 1965, p. 32), political participation can be incorporated as either inputs or demands into the political system. Milbrath (1977, p. 9) distinguishes between these inputs (i.e., campaigning, voting and protesting) and extractions (i.e., services, public order, and justice). While expressing their demands, citizens trust institutional settings to, at least, provide some basic common goods. The more they are able to control institutions or actors in said institutions, the better. In fact, even the possibility of political participation is able to restrain political actors: If decision makers believe that the ordinary man could participate [...] they are likely to behave quite differently." (Almond & Verba, 1989, p. 139). But citizens are not in any way frequent participants, they place checks on leaders at elections. They are 'controllers' rather than 'participants' (Parry, Moyser, & Day, 1992, p. 5). An active citizenry seems to be a myth of normative democratic theory. Citizens often lack the motivation to partake in political processes (Berelson, Lazarsfeld, & McPhee, 1954; Campbell, Converse, Miller, & Stokes, 1960; Neuman, 1986). Nonetheless, a democracy without citizen participation suffers the loss of its inputs on a systemic level. Furthermore, normative concerns cannot be completely ignored as Dalton stated: Without public involvement in the process, democracy loses both its legitimacy and its guiding force (Dalton, 2006, p. 35).

3 In contrast to elitist democratic theory (Sartori, 1962; Schumpeter, 1942), the participatory democratic theory underpins the importance of active citizens (Parry et al., 1992, p. 7). Yet empirical evidence shows that politics is not as relevant as other parts of everyday life (van Deth, 2000, p. 121). Dahl (Dahl, 1961) sets the task for political researchers to ascertain the question why few citizens are active, rather than focusing on elitists approaches. To grasp an understanding of political participation, it is necessary to distinguish it from other concepts. Citizen action might not need to be political in its nature. Concerning Easton (1953, p. 34), the act must have a relation to the authoritative allocation of values for a society. It is somewhat hard for citizens and researchers to assert precisely how a single action is linked to the mechanisms of a political system, thus restraining the scope of participation to mostly professional politicians. A broader approach by Verba (1972, p. 2) only focuses on legal activities by private citizens, a concept which became outdated in the participatory revolution of the 1960s and 1970s. The completely open definition by Milbrath (1977, p. 2) regarding all forms of participation to influence or to support government and politics includes a wide array of actions which are merely symbolic or relate to individual political interest and, therefore are not directly aimed at politics. Concerning the matter of political trust, knowledge, and participation, the citizens actions must aim at politics and be performed by individuals on their own account. For Kaase defined political participation as actions [ ] that citizens undertake voluntarily with the goal to influence decisions on various levels of the political system (Kaase, 1997, p. 160). Despite the definition being straight forward, political action seems an illusive construct to address. Over the years, modes to influence politics have evolved, widening the possibility and scope of involvement. Although directly aimed at politics, it is different to go to a campaign meeting rather than to an illegal demonstration, where one would hardly find the same people. Political participation is not one-dimensional (Gabriel, 2004; van Deth, 2009). What constitutes different domains of political participation is manifold. As Kaase (Kaase, 1997, pp ) stated, modes of political action can differ, if constitutionalized, in terms of legality, if a grade of legitimacy is achieved, or if they are conventional or unconventional (with Steinbrecher, 2009, p. 39 adding directness). These categories seem intuitive, but still are prone to discussion due to the fact of changes over time. Constitutionally guaranteed participation rights, like voting, or legal participation are not prone to change quickly, whereas legitimacy and conventions are more volatile: What might have appeared unconventional in the 1970`s [ ], today stand out as some of the most frequent used venues for citizen influence in Western Europe (Teorell, Torcal, & Montero, 2007, p. 343). If new participatory modes emerge, with internet activism being a contemporary example, it takes time for social conventions to change and a grade of legitimacy to be achieved. At first, few pioneers vanguard new modes of participation until they are adopted by a broader public. As Dahl (1961) said, it is not given that political involvement is a mass-phenomenon, and in this case, it can be disputed what a broader public constitutes, but nonetheless citizens use varying modes of political action to influence politics. Determinants of political participation are first foremost individual resources, political engagement and recruitment (Verba, Schlozman, & Brady, 2001, pp. 269ff). Although institutional settings may change the levels of political action (Verba & Nie, 1972), individual resources still play a major role influencing political action. As Barber put it: Any political action is a combination of opportunities and resources and motives (Barber, 1972, p. 12). Resources are merely described as socioeconomic factors, i.e., age, sex, and education, whereas motives are more closely linked to political interest, attitudes toward given objects, and even personality traits (for a comprehensive overview on classical modes see Milbrath (1977). One key factor is political efficacy. As internal political efficacy, the subjective assessment of one s own

4 individual resources relating to the ability to take action is linked to the likelihood of taking said action (Almond & Verba, 1989, p. 139). External political efficacy, on the other hand, is more closely linked to political trust as a feeling of responsiveness of the political system (Campbell, Gurin, & Miller, 1954; Vetter, 1997). Yet trust and efficacy are constructs on separate dimensions (Craig, Niemi, & Silver, 1990). First of all, internal efficacy is a subjective evaluation, and secondly, the feeling of responsiveness to one's actions does not imply effects on trusting political institutions. In its dichotomy, Milbrath (Milbrath & Goel, 1977, p. 70)displays their interaction. Table 1. Classification of political participation High High Active, Allegiant, and Conventional Participation Efficacy Low Supportive, Patriotic, and Ritualistic Participation Display according to Milbrath 1977, p 70. Trust Low Radical Action or Unconventional Participation Withdrawl from Politics Dividing the separate effects of trust and efficacy on different modes of political participation is reasonable enough, but it lacks a closer look on what efficacy is in the first place. Campbell and colleagues define efficacy as a feeling that political and social change is possible, and that the individual citizen can play a part in bringing about this change (Campbell et al., 1954, p. 157). This taps on evaluations of citizens own abilities. Later research sustains a two dimensional model, with internal efficacy in line with Campbell et al and external efficacy as a feeling of responsiveness by the political system to inputs by the citizenry (Vetter, 1997, p. 4) which seems closely related to political trust. Milbrath and Goeal (1977, p. 70) do not account for the two dimensions (i.e. they focus on internal political efficacy) and incorporates political knowledge in their concept of efficacy. Research on political knowledge and participation has stated that factual knowledge is indeed a separate factor influencing political participation (Barber, 1972; Gabriel, 2004; Johann, 2011, 2012). Barber (1972) even states: One of the key resources for political participation: knowledge (Barber, 1972, p. 12). Though Bathelt (2008) found that effects of political efficacy on political participation can be scarce for different modes of participation, Reichert (2016) states that the effects of political knowledge are mediated especially by internal political efficacy. The psychological foundation to address political problems and act accordingly seems to be political knowledge, especially with regard to complex forms of participation (Vetter & Maier, 2005, p. 52). Political Trust Political trust is a relevant factor for a working democracy. It safeguards a representative democracy (Gabriel & Zmerli, 2006; Zmerli, 2004), allows individual actions within an institutional setting, and is closely linked to the concept of political support (Fuchs, Gabriel, & Völkl, 2002; Gabriel, 2002). In the concept of political support, Easton (1975) distinguishes between diffuse and specific support. Diffuse support focuses either on an institution or on what it represents, whereas specific support refers to confidence in the perceived outputs and performance of political authorities. Political trust is, besides legitimacy, a key component of diffuse support (Easton, 1975, pp ). For a

5 democratic system to persist, it is necessary that at least moderate levels of diffuse support and, therefore, political trust are present. If a democratic state gains trust, it is able to act, to integrate and to perform (Gabriel, 2002, p. 496). Political trust in a democratic state is typically mimicked in classical models by the levels of trust in its institutions. Therefore, trust in institutions is essential for a systematic perspective (Offe, 1999; Warren, 1999). Furthermore, political trust is seen as having implications for citizens living in said democracy. It seems to heighten political involvement on an individual level (Almond & Verba, 1989; Mishler & Rose, 2005; Norris, 1999; Putnam, 1993, 2000; Uslaner, 2002). So, if changes in political trust occur, it not only influences general diffuse support criteria on a systematic level, but also affects individual inputs into the system by individual involvement. This adds to the problem of declining political trust in political institutions that most scholars have debated (Hetherington, 1998; Hibbing & Theiss-Morse, 1995; Rosenstone & Hansen, 1993) for institutions see (Miller, 1974a), for political actors see Citrin, 1974). Most scholars agree that "rather than simply reflecting dissatisfaction with incumbents and institutions, declining trust contributes to this dissatisfaction, creating an environment in which it is difficult for those in government to succeed" (Hetherington, 1998, p. 791). Political trust consists of manifold dimensions and interdependencies, and thus it is hard to define. According to Putnam (1993, p. 170), trust is an essential component of social capital. If interpersonal trust is in the center of trust relations it is questionable how institutions can acquire trust on a systemic level. Offe (1999) and Warren (1999) skeptically formulate an inductive spillover mechanism from interpersonal trust to institutional trust. Fuchs (2002, p. 427) on the other hand reformulating Easton (1975) states that trust is an attitude toward institutions similar to legitimacy of political institutions. He also states that institutional trust is better off as a conceptual dimension of political support rather than social capital (for a full discussion and empirical results see Fuchs et al., 2002). Defining trust as an organizational concept can be seen as "the willingness of a party to be vulnerable to the actions of another party based on the expectation that the other will perform a particular action important to the trustor, irrespective of the ability to monitor or control that other party" (Mayer, Davis, & Schoorman, 1995, p. 712). Transferring this concept to political trust can be achieved by adapting the view that citizens are always vulnerable to the actions of their governance. Therefore, institutions directly matter for their everyday life (for a similar approach defining trust in politicians see Halmburger, Rothmund, Baumert, & Maier, 2015, p. 5). Citizens are occasionally able to monitor and control actors in governments by casting their vote. Of course, this form of direct control does not apply to all political institutions in a given system. It rather depends on the systematic setting of the specific nation. Institutional trust is even more important concerning that institutions are much more difficult to control than specific politicians, one cannot vote for or against them. Following this argument, trust in institutions is better off as a separate dimension in the concept of diffuse political support (Fuchs et al., 2002, p. 438), rather than as a derivative of aggregated interpersonal trust. On an individual level, political trust constitutes an attitude towards institutions (Miller, 1974a, 1974b; Stokes, 1962), which can be distinguished from other attitudinal constructs, such as political efficacy (Craig et al., 1990). Political Trust and Participation If political trust is a distinctive attitudinal dimension the question arises: Has trust meaningful effects on other variables (Hetherington, 1998, p. 792)? Research states that higher levels of political

6 trust further democratic values (Norris, 1999) and lead to the rejection of "undemocratic alternatives" (Mishler & Rose, 2005, p. 1053), see also Rose, Mishler, & Haerpfer, 1998). Empirical findings underpin the theoretical implications described above, trust seems to increase support (Brehm & Rahn, 1997). Also, political involvement seems to be dependent of political trust (Mishler & Rose, 2005; Norris, 1999; Putnam, 1993, 2000; Uslaner, 2002). Generalized political trust can predict voter turnout and it seems that high levels of generalized trust lessen the effect of specific politicians on voters choice (Halmburger et al., 2015, p. 3). Not only voting as a low threshold type of participation (Vetter & Maier, 2005, p. 52), but every form of political participation seems to be influenced by political trust (Milbrath & Goel, 1977, p. 70). Political Knowledge Citizens ability to follow, remain attached to, and even getting involved in politics is often discussed given that just moderate levels of political knowledge are present in the general public. Converse (Converse, 1990, p. 372) states that the two simplest truths I know about the distribution of political information in modern electorates are that the mean is low and the variance is high." In their groundbreaking study on political knowledge Delli Carpini and Keeter (1996, p. 1) highlighted that, for democratic systems to function, an informed public is needed, but do not share Converse's concerns with regard to knowledge being scarce. The functionality and stability of a democratic system require its citizens to know about its setting at least. Immigration to Germany, for example, necessitates courses on political knowledge in the naturalization process (Bundesministerium des Inneren, 2008, p. 144). The normative relevance of politically informed and therefore competent citizens is often asserted (Bennett, 1989; Delli Carpini & Keeter, 1996; Neuman, 1986, p. 1; Thompson, 1970, pp ). The public good of political knowledge derives from the idea that informed citizens are able to evaluate their own interests and act accordingly, which in turn underpins a normative notion of the participatory democratic theory: It is a central premise of democratic practice that the citizenry must be both vigilant and vocal about issues which affect them if the system is to work as intended. (Neuman, 1986, p. 1), see also (Delli Carpini & Keeter, 1996, p. 7) Interestingly enough, even the general public seems to realize that they should track information about politics (Bennett, 1989, p. 422). Delli Carpini and Keeter (1996, pp ) consider the knowledge of issues, political actors, and the institutional setting as indispensable for any citizenry and society. It enables citizens to act in their own interests in political situations. Therefore, higher aggregate levels of knowledge empower societies to be more inclusive and responsive. Additionally, voters are enabled to fulfill their role as controllers in the political system when making informed decisions (Dalton, 2006, p. 13; Westle, 2005, p. 484). Concerning the wide range of information available in a democratic system, the question arises as to what citizens need to know. At least, some kind of understanding of how a democracy works has to be acquired, or as Barber put it: "Supposedly a citizen would need to know what the government is and does before he could make a rational judgment about political questions" (Barber, 1972, p. 44). This focuses on institutional settings as well as on relevant issues, but in the case of voting, the decision to make is generally to chose between politicians. Delli Carpini and Keeter add that some knowledge about political actors has to be acquired as well: these three broad areas what we call the rules of the game, the substance of politics, and people and parties provide reasonable organizing principles for discussing what people should know about politics (Delli Carpini & Keeter, 1996, p. 65).

7 One concern about what citizens should know regards the differentiation between what people think they know and what actual level of information can be attested: objective political knowledge is a subset of subjective political knowledge (Maier, 2000b, p. 143) and personal involvement in politics, too. To gauge levels of actual political knowledge, the definition as the range of factual information about politics that is stored in long term memory" (Delli Carpini & Keeter, 1996, p. 10) seems appropriate in order to derive valid measurements of knowledge. If the salience of politics is rather low for the common citizens, it is questionable if they remain informed about each part in polity, politics, and policies (McGraw & Pinney, 1990). This would not be a problem if there were not presumably different effects of generalized versus domain-specific knowledge. The discussion whether political knowledge is a general trait (Delli Carpini & Keeter, 1996, p. 136; Neuman, 1986; Zaller, 1986, p. 1) or citizens acquire information specifically in given subsets (Bennett, 1990; Owen & Stewart, 1987) is ongoing. Iyengar (Iyengar, 1986, 1990, p. 166) provides valid evidence that knowledge is domain-specific, and recent studies provide us with results pointing in that direction as well (Westle & Johann, 2010). Changing modes of information transmission might turn the citizenry into information specialists (Kim, 2009, p. 255) even more, which is of concern when investigating the effects of specific political knowledge. Political Knowledge and Participation Individual political action, as shown above, depends on one s resources, motives, and opportunities. Can political knowledge be a part of one of these dimensions? Neuman (1986, p. 5) suggests that the measurement of socioeconomic resources and attitudes is not of great importance, but the central issue is the correlation between political knowledge and either opinion or behavior (Neuman, 1986, p. 5). A precondition to this analytic strategy must be the circumstance that political knowledge, on its own account, can affect political behavior. Research suggests that effects of factual information on political action are substantial (Barber, 1972, p. 12; Gabriel, 2004; Johann, 2011; Junn, 1991; Leighley, 1991). In the well documented case of political turnout, high levels of political knowledge increase the turnout (Delli Carpini & Keeter, 1996, p. 225; Larcinese, 2007; Neuman, 1986; Wattenberg, McAllister, & Salvanto, 2000; Westle, 2009), yet Howe (Howe, 2006) shows that the effect must be linked to the institutional and social settings (in Canada, it seems to be the trend that knowledge promotes turnout, whereas in the Netherlands it does not) and Reichert (2016) states that knowledge has no direct effect on conventional participation. More complex and closed types of political participation also seem to be affected by one s political competence. This can be due to the fact that a knowledgeable citizen simply has the information about possible modes of participation (Lemert, 1981; McClurg, 2006; Milner, 2002; Popkin & Dimock, 1999; Verba, Burns, & Schlozman, 1997; Verba, Schlozman, & Brady, 2001). Knowledge promotes non-electoral participation (Delli Carpini & Keeter, 1996), even divided into conventional and unconventional (Westle, 2012, p. 62). In relation to political involvement, a multidimensional approach is achieved if we assume that categories of political knowledge (for example issues or actors) affect political action in different ways. Johann (Johann, 2011, pp , 2012) provides evidence that if we separate modes of political participation we also have to take a look at the different domains of political knowledge. Knowledge of different domains has different effects, for instance, the knowledge of political actors affects voting, whereas information on polity is more relevant for other modes of participation like protest. Connecting Knowledge, Efficacy, Trust, and Participation The relation between the relevant constructs information, efficacy, trust, and action is neither imminent in political theory, nor is it obvious in empirical research. Neuman states that political sophistication [with political knowledge] is strongly correlated with political activity (Newman

8 1986, p. 83), but sophistication does not cause participation. It functions instead as an intervening variable (ibid., p. 103; Vetter & Maier, 2005). In addition, as described above, research shows that knowing the modes of participation which are available, i.e., knowing about polities, surely has a positive impact on types of participation more complex than voting. Also, political action does not only depend on the opportunity to act but also on individual resources and motives. Political knowledge can be seen as a resource, in terms of it being derived from political socialization and education. Citizens must have personal resources in order to act politically. For example formal education must not be the predictor of political action, it could be political knowledge accumulated in education. As motive, first individual political interest is a key predictor to get involved in politics. However, before action is taken, one has to know how to get involved. Also knowledge connects attitudes to political decision (Alvarez, 1998) then again, one's attitudes regarding politics can be shielded against persuasion by political knowledge (Delli Carpini & Keeter, 1996; Kinder & Sanders, 1990). If political efficacy is an attitude toward the political system (external) and an understanding of one s own abilities (internal), political knowledge can in fact mediate the process in which decisions to act politically are made. Reichert (2016) makes the case that action theory suggests that subjective feelings of political efficacy might be more relevant than factual knowledge when predicting actual behavior. On the other hand, [research] also suggest that political knowledge be a vital determinant of political efficacy (Reichert, 2016, p. 221) If citizens think that their actions cannot influence politics in general, they might know the rare possibilities when participation does matter. If one considers the democratic system as irresponsive, unconventional action might be preferred, with the need to know about these possibilities. Milbrath's (1977, p. 70) table for different levels of trust and efficacy is a starting point in order to grasp an understanding of the various underlying effects between all these constructs. An assumption can be made that all three constructs are indeed correlated, with internal efficacy being closer to political knowledge (Reichert, 2016; Vetter & Maier, 2005) and political trust being closely related to external efficacy. Surely, political trust and political efficacy may have different effects on different modes of participation and it is also questionable how political knowledge affects external efficacy and political trust. As an example, if voters think that the political system is responsive they are more likely to vote in general. This effect could be even stronger if they do not trust political institutions, since they exercise their voting rights more frequently, acting as controllers for political actors, and therefore covering the basic problem of political trust (as stated in the definition). Yet this requires an understanding of the political processes, i.e. knowledge about polity. So it is possible that effects of external efficacy on political behavior are mediated by political knowledge or vice versa when a general understanding of the political process generates political trust and or efficacy. As the set task of this paper is to tap on these problems it is necessary to address these questions in terms of predictive power of different modeling approaches. Therefore, I do not formulate any hypotheses, but instead set out to empirically explore the two research questions at hand: RQ 1: Has political knowledge a distinct effect on different modes of political participation when controlled for known predictors? RQ 2: Is political knowledge a mediator in models of political participation or is political knowledge mediated by other predictors like efficacy or trust?

9 Data and Methods To analyze the knowledge, trust and participation a data-set is needed in which all three constructs are measured simultaneously. The ALLBUS provides such data. The biyearly conducted survey for Germany consists of personal interviews with standardized questionnaires (CAPI) including a battery of political knowledge administered as a computer assisted self-interviewing process to minimize interviewer effects. Additional 3,469 interviews were conducted in 111 primary sampling units (communities) in West Germany and 51 in East Germany. The response rate was 40.5 % (west) and 39.5 % (east) for a combined total of 40.3 %. To prepare the data for analysis respondents with overall missing values on relevent questions are eliminated, leaving 2,797 (west) and 636 (east) for a N of 3,433 (GESIS - Leibniz Institute for the Social Sciences, 2011). Operationalization of Key Variables Political trust is measured by a battery of items ranging from government to public media on a seven point scale from 1 = no trust at all to 7 = very high trust. The three items, used in these analyses to measure trust, are the questions about parliament, government, and political parties. These institutions display the trust in the political system best (Fuchs et al., 2002). Two indices are constructed to measure internal and external efficacy, both consisting of two items. I measured internal efficacy by the answers I can be active in a political group and politics is too complex for me, external efficacy by Politicians advocate the citizens interests and Politicians do not care about what people like me think (Vetter, 1997). Political knowledge is measured as a series of factual questions. Eight questions about political actors are selected, in which respondents had to connect politicians to their party or name the right politician on a given position as an index of knowledge for who government is. Six polity questions and one for policy are connected to an index on what government does and how it works. A combined knowledge index uses all 15 items. The indices are coded as percentage of right answers from 0 to 1. All predictor variables (sex, party identification, political interest, education, age, internal efficacy, external efficacy, and political trust are either dummy variables or coded as percentage between 0 and 1 in order to achieve comparability. The predictors are always coded as 1 with the highest value, meaning that respondents trust institutions, feel that the system is responsive, see themselves as competent, and know more about politics if the values reach 1 on the scale. Table 2: Overview of indices trust internal efficacy external efficacy knowledge about polity and policies knowledge about actors knowledge combined Mean Std.Dev N Political participation consists of twelve items (administered as a card game to reduce interviewer effects) of possible participation. Respondents are asked if they would use said mode of participation 2. Two participation modes were excluded: first, say political opinion to friends or coworkers, and second, political consume as buying goods under political, ethical, or environmental 1 ALLBUS Data can be retrieved from 2 This paper uses would participate, instead of had participated (a follow-up battery), because retrospective participation must not depend on actual present attitudes.

10 considerations. These two modes do not meet our definition of political participation, in that they are not directly aimed at influencing the political sphere. Methods In a first step, it has to be established which predictor shows a significant influence on the different modes of participation. Therefore unconditional binary logit regression models are used to grasp an understanding of how predictors act under control of the whole set of predictors. This is a conservative approach to addressing the power of a predictor, since I use only predictors with significant coefficients in later analysis. For every mode of participation two regression models are calculated, one with a differentiation between knowledge domains and one with the combined knowledge index to address the possible distinction between effects of specific knowledge (Johann, 2011, 2012). In a second step predictors that have significant effects are selected for a path analysis, due to the premise that predictors and mediators must have significant effects in order to start a path analysis. For every mode of participation, significant predictors are selected due to the findings in the preceding logit models and simple mediator models are conducted. For every predictor, two path models are calculated to address the question of the position of political knowledge in the model. Figure 1. Logic of Path Analyses for simple mediator models The second path analyses are conducted only if there is a theoretical argument for a mediation to happen (for example, political knowledge will not influence age or gender). Additionally predictors with no influence on political knowledge are not used in path models since there no is theoretical assumption in the classical view on these models that mediation can happen if there is no connection between predictor and mediator (Hayes, 2013; Kenny, 2014). Bear in mind that Kenny (2014) suggests that the total effect cannot be simply computed as c +ab if the dependent variable has a dichotomous outcome, yet Hayes (Hayes, 2013) points out that his PROCESS-macro can incorporate variables with a dichotomous outcome. To be able to compare the results, I chose to conduct a simple mediator model 3, as described above, for every predictor (coded to the same range) and not to incorporate additional predictors into the model directly. If different predictors influence others, this would be seen in the binary logit models of step one. If there are predictors which are not significant, they will not be used in the path analyses, hence the conservative approach. 3 Computed in SPSS by bootstrapping with PROCESS-macro v2.13 (

11 Results Reviewing the effect sizes of different predictors as they are calculated in a complete binary logit models addresses our first research problem. Reichert (2016) discussed the issue at hand and found no distinct influences of political knowledge when controlled for other predictors. A closer look on the effect sizes in Table 3 provides insight into the question how political knowledge affects political behavior when different modes of participation are examined separately. At first glance one can state, that political knowledge has a significant effect on modes on behavior even controlled for other predictors. Also political attitudes, normally connected with political involvement, do not eradicate the effect of political knowledge. Yet it is key to distinguish between different domains of political knowledge as Johann (2011) stated as well as different modes of participation (Milbrath & Goel, 1977) since effect sizes and orientations of all differ to a great extent between modes. In the case of voting, respondents are more inclined to the intention to vote if overall levels of political knowledge, external efficacy, political interest, and political trust are high. Yet voting does not seem to be dependent of internal political efficacy. This is an indication that citizens are well aware of the low threshold more supportive act as that voting can be described. The occasional voting for another party is a completely different case. Here better informed citizens use this tool for the control of the governmental elites which they do not trust and a system which does not seem responsive to them. To abstain from voting is largely dependent of distrust but higher levels of political knowledge on the functioning of a political system prevent citizens from absenteeism. If connected to the categories by Milbrath and Goel (1977) we might find that voting is more supportive and ritualistic and nonvoting describes the withdrawal from politics, which indicates a quite accurate picture. Solely the inclination to vote for another party would be considered as an unconventional mode of participation by this classification, a classification which is a bit out of focus even in a 2 by 2 typology. Concerning forms of unconventional participation the classification table does not provide enough discriminatory power. Surely illegal demonstrations are to classify as unconventional participation or radical action largely dependent on low levels of political trust and a higher internal efficacy but not on political knowledge. Yet legal demonstrations and online protest, in the last century also forms of unconventional action, must be switched into active participation with dependant on higher levels of trust, internal efficacy, and political interest. These modes of behavior are also dependant on higher levels of information about polity whereas for illegal demonstrations factual information does not change the inclination to participate. Public discussions and signing a petition may have also been considered unconventional at times, yet as low threshold modes citizens as well as political systems are accustomed to these forms of input. In their dependencies form our predictors both modes show some differences. Political interest and internal efficacy seem to promote these types of behavior, yet public discussions are dependent on knowledge about the system whereas signing a petition profits from a general range of factual information. Externals efficacy and trust does not provide any significant effect on these modes rendering the classical typology obsolete in locating these types of behavior. Working in parties and civic action groups could be describes as active conventional participation. The knowledge effect is scarce for civic action groups, inhibiting a further analysis in our path models, but effects of internal efficacy and political interest are positive significant with no effect of political trust. The traditional work in a political party on the other hand is negative dependent of political knowledge. This interesting finding means that respondents with lesser factual information about the political sphere in general are less inclined to work in the political system. All other predictors display the expected directions and effects, higher levels of political efficacy, interest, and trust promote working in parties classifying this mode as classic conventional active participation.

12 Table 3. Effects of Knowledge on Partizipation Voting Voting for Other Party Nonvoting Legal Demonstrations Illegal Demonstrations Knowledge of B SE N.R² B SE N.R² B SE N.R² B SE N.R² B SE N.R² Polity *** *** Knowledge Politicians of Combined *** *** inteff * ** *.370 exteff.684 * ** trust ** ** *** ***.494 int.979 *** * ** *** **.369 Standard PID.725 *** * * ** predictors ow.417 ** * *** * sex *** *** *** age *** *** *** ***.501 edu *** *** Online Protest Public Discussion Signed Petition Civic Action Group Work in Party Knowledge of B SE N.R² B SE N.R² B SE N.R² B SE N.R² B SE N.R² Polity *** ** Knowledge Politicians of Combined *** * inteff *** *** * *** ***.291 exteff *.319 trust **.373 int.599 * *** *** *** ***.300 Standard PID.255 * * *** **.126 predictors ow.359 * * * sex * *** *** age *** *** *** * ***.368 edu.788 *** * ** Significance Levels: ***=p<.001. **=p<.01. *=p<.05

13 In sum the findings of the binary logit models yield some interesting results. First it has to be stated, that the classification table by Milbrath and Goel (1977), despite its age, works reasonably well. Some forms of participation does not fit the descriptions and presumably modes previously described as new or unconventional are now part of the participatory canon (Teorell et al., 2007), the general idea of the typology although still holds. Not displayed in a 4 by 4 table are the different effects of the dimensions of political efficacy. So in all models (except for public discussions) external efficacy is exactly aligned with political trust. It can be assumed that in some cases political trust as a control variable renders effects of external efficacy insignificant. Internal efficacy is more closely related to political interest and political knowledge, and in fact displays the dominant idea behind the typology. The case of political information yet remains not as clear. Is it a part of internal political efficacy, is it mediated by internal efficacy, or are there unitary effects of factual information? In the cases of voting and voting for another party factual information provides an independent effect. For other modes, despite illegal demonstrations and civic action groups, effect sizes stay visible under control of the whole set of predictors (internal efficacy included), which indicates unitary effects of information on political behavior. Unclear remains the possible mediation between the predictors added in the models. Mediation effects of political knowledge Possible mediation effects are hard to grasp for different modes of participation and different predictors at the same time, especially if one wants to compare the models and decide which suits the data best. Therefore each model must be calculated separately and with the same setting with at one time knowledge as a mediation variable, and a second time knowledge as a predictor. Therefore, in tables 4 and 5 I will show an example of the different possible interactions between knowledge and other predictors for signing petitions and public discussions. For additional information on the other modes ill provide an overview of the interpretation in table 6 and full information in the appendix. In starting to understand the different relations underlying political action and its predictor s one can derive from effect sizes of the binary regression models. The process of mediation however similar in approach doesn t provide us with a single effect size for interpretation rather than four effect sizes in one model and an overall total effect. In this modeling, the total effect can be seen as applied without moderation. An added moderator should decrease the total effect when the model is computed. In my example (signing a petition) this is the case for internal efficacy, but not for political interest. Therefore, moderation does not happen through knowledge as far as interest is concerned, but moderation does happen with internal efficacy. For a second step, one has to take a look at the second part of the table in order to compare effects of knowledge as the first predictor in the model. Here we see that the effect of knowledge is mediated by both internal efficacy and political interest. This raises the question which form of causal assumption fits the data better. If indirect effect sizes are compared, knowledge as mediator performs better for internal political efficacy; additionally the direct effect of political knowledge under control of internal efficacy is better suited to influence political behavior in this case. It can be assumed that political knowledge functions as a mediator of internal efficacy when signing a petition is concerned. Political interest, on the other hand, constitutes as a mediator of political knowledge since the total effect is only moderated in the last line of table 4. Party identification, gender and age are not included in the second step since they are either dichotomous variables (pid, gender), not suited for serving as a mediator or not prone to be changed by political knowledge (gender, age).

14 Table 4. Signing a Petition Total Effects (c) Direct Effects (c') Indirect Effects (ab) Effect X M Effect M Participation Knowledgeindex: Combined effect Sign. SE effect Sign. SE effect Sign. Boot SE effect Sign. SE effect Sign. SE Mediator (M) Predictor (X) internal efficacy *** *** *** *** ***.152 political interest.534 *** *** *** *** ***.158 party identification.469 *** * *** *** ***.144 gender * *** *** *** *.071 age *** *** *** *** ***.147 internal efficacy *** *** *** *** ***.173 political interest *** *** *** *** ***.170 Significance Levels: ***=p<.001. **=p<.01. *=p<.05. For indirect effects, p based on normal theory tests. Table 5. Public Discussion Total Effects (c) Direct Effects (c') Indirect Effects (ab) Effect X M Effect M Participation Knowledgeindex: Polity effect Sign. SE effect Sign. SE effect Sign. Boot SE effect Sign. SE effect Sign. SE Mediator (M) Predictor (X) internal efficacy *** *** ** *** **.147 political Interest *** *** *** *** ***.173 party identification.457 *** *** *** *** ***.135 gender.211 * *** *** ***.137 age *** *** * * ***.135 internal efficacy *** ** *** *** ***.173 political Interest *** ** *** *** ***.170 Significance Levels: ***=p<.001. **=p<.01. *=p<.05. For indirect effects, p based on normal theory tests.

15 A different picture can be derived from the analysis for public discussions. Here, mediation does happen across the board and especially for internal efficacy and political interest. Comparing indirect effects between the two steps highlights that moderation happens when knowledge is set as a predictor which can be moderated by internal efficacy and political interest. Also, the effect sizes of these predictors on participation under control of knowledge suggest that this is the case here. For the overall model we see that a great deal of information has to be considered for the interpretation of these models. Table 6 aims at simplifying this task in providing an overview for every mode of participation in terms of knowledge being a mediator or a predictor, derived from the information in appendix a mainly focusing on the size of indirect effects. Table 6. Overview of mediation model comparison internal efficacy external efficacy political trust political interest Signing Petition M - - X Public Discussion X - - X Voting - M 0 M Voting for other party - M 0 M Nonvoting 0 - X X Legal Demonstration X - - M Online Protest X - - M Work in Party X X 0 X M = Knowledge is a mediator; X = knowledge is mediated; 0 = no mediation but independent effects; - = not computed since predictors not significant in binary logit regression The findings on mediation do not show a clear picture. In the case of casting ones vote even for another party, political knowledge mediates intentions to behavior for external efficacy a. Nonvoting is the only case where knowledge is mediated by political trust, and political trust seems overall independent from influences of political knowledge, as well as the other way around. External efficacy mediates the knowledge when it comes to working in a political party and internal efficacy is mostly mediating political knowledge. For political interest the mediation process is split by the types of political action, therefore it doesn t provide us with insight on how to address this phenomenon. Summary and discussion To sum up our findings so far is a daunting task. Generally, the idea holds that political knowledge has distinctive effects on political participation on its own account. Controlled for other predictors it holds a significant effect for almost all modes of political participation (with the exception of illegal demonstration and working in civic action groups). Interestingly enough, knowledge has a weaker effect when levels of political trust and external efficacy yield high impacts on political participation. In cases combined with internal efficacy and political interest better predictive power can be achieved. Also one has to account for different domains (Johann, 2011) of political knowledge, where different modes of political participation are concerned since effect directions can be contradictory. Overall, combined knowledge works well for low threshold participation (voting or signing petitions), for more taxing modes (demonstration or protest) knowledge of the system is important. The first research question - has political knowledge a distinct effect on different modes of political participation when controlled for known predictors can be answered with a yes.