Political knowledge of primary school children

Transcription

1 Political knowledge of primary school children Paper prepared for delivery at the IPSA World Congress, Montréal 2014 Panel Political socialization of children Abstract: Only on the basis of information and political knowledge can citizens form an independent opinion and competently participate in political decision-making processes. This paper investigates how and why political knowledge varies among young children. The data of the following empirical analyses is adapted from the German study Learning to Live Democracy (LLD). Therefore, the data used here consists of a three-wave panel of about 750 young children who were interviewed at the beginning and at the end of their first school year, as well as at the end of their fourth school year. The empirical results show that political knowledge among young children is not equally distributed: girls, children from Turkish families and/or those from areas with a low socioeconomic status show the tendency to be less politically knowledgeable than boys, children from native German families and/or those from areas with a higher socioeconomic status. These differences do not disappear during the first years in school. Apparently, school attendance is not enough to compensate for the initial inequalities. 6/27/2014 Simone Abendschön Department of Social Sciences Goethe-University of Frankfurt, Germany Abendschoen@soz.uni-frankfurt.de Markus Tausendpfund Department of Political Science University of Hagen, Germany Markus.Tausendpfund@fernuni-hagen.de

2 1. Introduction Political knowledge is a powerful tool for understanding the world. Only on the basis of political information and knowledge can citizens form an independent opinion and competently participate in political decision-making processes. Moreover, political knowledge is a powerful predictor of political participation and promotes support for democratic values. The unequal distribution of political knowledge throughout the population is rooted in three factors: ability, motivation, and opportunity (Delli Carpini & Keeter, 1996, p. 106). All of these factors, so we argue, are developing as early as childhood, as part of political socialization processes continue to thrive in the educational and civic careers of individuals. However, there are hardly any up-to-date findings regarding political knowledge gaps in childhood. The discussion of political knowledge gaps concentrates almost exclusively on adults (lack of) political knowledge and the role that resources play in this regard. As important as such analyses are, they neglect, in our view, the fact that an individual s political knowledge is acquired through political socialization and learning processes which begin as early as childhood. International programs such as CivEd and ICCS are important cross-national sources of civic education and civic literacy and have recently revitalized research on 14-year-olds political orientations and civic education. This neglects the fact that political socialization and civic learning starts much earlier in a citizen s biography. Since the political inclusion of children is increasingly seen as a necessary condition in democracies and since findings from developmental psychology suggest the relevance of political socialization processes in childhood, the need for further research on younger children becomes evident. With our paper, we d therefore like to introduce a childhood perspective on the topic of resource-based political knowledge gaps. We d like to investigate if, how and why political knowledge varies among young children in primary school. Are there already significant gaps in political knowledge among young citizens? When do differences in levels of political knowledge become visible? And how can knowledge gaps possibly be explained? The empirical analyses draw on panel data from the German project Learning to Live Democracy. 1 The study includes a three-wave panel of about 750 young children who were interviewed at the beginning and at the end of their first school year, as well as at the end of their fourth school year. The children were questioned with a specially developed standardized questionnaire which included, among others, items regarding their political knowledge, as well as their recreational activities (e.g. media use). All three waves are analyzed in this paper. This three-wavepanel-perspective allows for a detailed description and analysis of political knowledge, its development and explanation at a young age. After a brief discussion and the development of the role and concept of children s political knowledge, the paper goes on to develop hypotheses for children's political knowledge gaps. It will 1 The study was initiated in 1999 at the Mannheim Centre for European Social Research by Jan W. van Deth, who has directed the project since its inception. From 2004 to 2007, the German Science Foundation (DFG) has completely financed this study (Grants DE 630/11-1 and 11-2). This support is gratefully acknowledged here. For more information on the study, see (accessed 27 June 2014). 1

3 then empirically analyze children s political knowledge and potential inequalities. We will also discuss and examine the development of children s political knowledge (gaps) during their primary school career. 2. Political knowledge and children From the perspective of democratic theory, politically well-informed citizens are quite desirable. Political knowledge is a prerequisite for the perception, evaluation and understanding of democratic political life. It is a well-established fact in political behavior research that being politically knowledgeable fosters and facilitates political participation, it delivers an informed background to political decisions and it enables people to recognize and articulate their political interests (Delli Carpini & Keeter, 1996; Niemi & Junn, 1998; Milner, 2002). Political knowledge is positively related to other political orientations such as political efficacy, trust (Galston, 2001), and tolerant attitudes (e.g. Dassonneville, Quintelier, Hooghe, & Claes, 2012). Consequently, it is considered the currency of citizenship (Delli Carpini & Keeter, 1996, p. 8). Empirical studies, however, repeatedly find rather low political knowledge levels among citizens (e.g. Maier, 2009; Converse, 1990). Generally, citizens cannot live up to the democratic ideal of the knowledgeable homo politicus who displays consistent and stable political attitudes. That some citizens know more than others is however only problematic if different knowledge levels are not arbitrarily distributed throughout the population, but can be linked to specific structural characteristics and, therefore, parts of the population display systematic and permanent knowledge deficits (e.g. Westle, 2011). These systematic inequalities have been demonstrated in many studies with adults (e.g. Luskin, 1990; Elo & Rapeli, 2010; Glantz, Bathelt, & Maier, 2012). They are rather worrisome from a democratic viewpoint because they lead to unequal participation and representation in political life (Westle, 2011, p. 835). Whereas these facts are rather undisputed in political knowledge research, aspects of the conceptualization of political knowledge are more under discussion. There are related definitions to political knowledge such as political sophistication (Luskin, 1987), political awareness (Zaller, 1992) and many others (for a detailed discussion of these concepts see for example Lawrence, 2003, pp. 7-22). Delli Carpini s and Keeter s definition of political knowledge as the range of factual information about politics that is stored in long-term memory (Delli Carpini & Keeter, 1996, p. 10) seems a central feature to most of them. This factual information should be correct (Luskin, 1987) or be proven to be true (Dekker & Nuus, 2007, p. 28). These basic approaches shall also suffice as a starting point for our analysis of children s political knowledge, for which there exist, at least to our knowledge, no special concepts so far (see alsovan Deth, Abendschön, & Vollmar, 2011; Vollmar, 2012). Neuman s (1986) conceptual and empirical distinction between three dimensions of political sophistication might additionally be helpful for our study of children s political knowledge. He distinguishes between political salience, political knowledge and political conceptualization. According to Neuman, political salience is the initial component of political sophistication. It 2

4 encompasses interest, concern, and attentiveness towards politics. The second and core component is political knowledge, which Neuman defines as familiarity with major political issues and accurate knowledge of prominent political figures and events. There is a consensus that the core of individual knowledge should contain knowledge of what the government is and does (Barber, 1969, p. 38; Delli Carpini & Keeter, 1993, p. 1182; Neuman, 1986, p. 196). The individual should, therefore, be familiar with the basic structure of government, prominent political actors and current events (Neuman, 1986, p. 196). The last and most demanding aspect of political sophistication is represented by the political conceptualization dimension. This is, according to Neuman, the cognitive organization by means of abstract concepts and the active use of political knowledge in the evaluation of political issues. Although Neuman states that these three dimensions should not be understood as stage or sequential models of cognitive development (in the tradition of Piaget or Kohlberg), he plausibly suggests that political attentiveness may lead to the gradual accumulation of a reservoir of factual knowledge, which in turn manifests itself in political conceptualization (Neuman, 1986, p. 54). Therefore, we think that political salience and factual political knowledge may be a good conceptual and empirical start to study children s political knowledge with the data at hand. Early political socialization studies (Hess & Torney, 1967; Greenstein, 1965; Easton & Dennis, 1969) have already indicated that the acquisition of political facts and information starts in childhood. Children acquire political knowledge, next to other political orientations, during political socialization processes. Although the focus of these early studies was mainly on affective attitudes and political support how can political systems reproduce themselves? they provided first valuable insights at the beginning of political learning. They showed that, during the primary school 2 years, children acquire more and more political information (Easton & Dennis, 1969; Hess & Torney, 1967; Moore, Lare, & Wagner, 1985). 3 More recent studies also show that primary school pupils are politically knowledgeable (e.g. Gemmeke, 1998; Hafner, 2006; Götzmann, 2008; van Deth et al., 2011). This means that regarding the development of children s political knowledge during their four years in primary school, we expect that HL 4 1: the political knowledge levels of children should rise from wave to wave. 3. The usual suspects and socialization influences: Sources of political knowledge gaps at a young age Political learning in childhood, and thereby the acquisition of political knowledge, is influenced by the child s surrounding socialization contexts; at a young age mainly the family, followed by school and peers. Nowadays, the media also plays an important role regarding political learning (van Deth et al., 2011; Bronstein, Daily, & Horowitz, 1993). As socialization studies show, these socialization contexts 2 The term primary school ( Grundschule in German) needs some clarification in the German educational context. Grundschule encompasses grades 1 to 4 (in the German federal states Berlin and Brandenburg grade 1 to 6). Children start grade 1 normally aged 6/7, secondary schooling starts with grade 5. 3 This knowledge centered especially on the person and the office of the president. 4 The data of the children s study we use in this paper allow both for a cross-sectional and a longitudinal perspective on children s political knowledge. Therefore we will formulate expectations/hypotheses that refer to both perspectives. The denotation H refers to a crosssectional; HL to a panel perspective. 3

5 are also heavily dependent on socio-structural characteristics of the child s family background. Although there are hints at socio-structural inequalities in the older studies, these have not been studied systematically recently. Therefore, as has been found with adults, the usual suspects can be assumed to contribute to the political knowledge levels of children as well. In the following, we will discuss the factors which might contribute to political knowledge gaps among children. 3.1 Socio-structural factors In terms of socio-structural characteristics, we distinguish between three categories: gender, age and socioeconomic background. Gender Women have repeatedly been found to expose lower levels of political knowledge than men (e.g. Dow, 2009). There are several explanations for this gap; two of the most common explanations can be termed the situation hypothesis and the socialization hypothesis (Westle & Schoen, 2002). Until a few decades ago, women used to be less educated than men and spent more time committed to family and (child) care. Therefore, they are, according to the situation hypothesis, supposed to be less knowledgeable than men (see for instance Burns, Schlozman, & Verba, 2001; Delli Carpini & Keeter, 1996; Ferrin, Fraile, & Rubal, 2012; Fraile, 2014). From this perspective, one would expect that there is no difference between girls and boys regarding their political knowledge (Ferrin et al., 2012). However, the second explanation attributes the gender gap in political knowledge to socialization processes which are different for boys and girls. It ascribes children s gender differences usually to the learning of different gender roles (cf. Orum, Cohen, Grassmuch, & Orum, 1974, p. 198). Girls are raised to be more empathetic, caring and home-oriented than boys. 5 The political knowledge of children then reflects gender differences found among adults; boys are often better informed about politics than girls (cf. among others Haug, 2013; Dowse & Hughes, 1971, p. 55; Greenstein, 1965, p. 116; Moore et al., 1985, pp ). Simply spoken, boys are always more political (Greenstein, 1961, p. 360). These gender roles could already be effective at the young age of the children studied here. From this perspective, we would expect our analysis to show that: H1a: Girls display significant lower levels of political knowledge than boys. As has been recently suggested by other students of political knowledge, knowledge and therefore also knowledge gaps can differ across political domains and areas (see Dolan, 2011; Ferrin et al., 2012; Stolle & Gidengil, 2010; Shaker, 2012; Fraile, 2014). This brings to mind older arguments of feminist theory stating that women do not know less about politics or are not less interested in politics than men, but that they might be less knowledgeable and interested regarding the predominant male concept of politics with its focus on established politics (e.g. Meyer, 1992). These different political 5 There are several other explanatory approaches for gender differences in political knowledge and/or political involvement in general, most noteworthy maybe methodological explanations (see especially Mondak & Anderson, 2004) which attribute a substantial part of the gender differences to question formats and response sets in representative surveys. 4

6 preferences might also be the case for young boys and girls. Therefore, we formulate an additional hypothesis: H1b: The gender differences may not occur throughout all knowledge domains. The start of the school career represents an important event in early childhood, which implies considerable changes and the widening of a child s world (e.g. Hurrelmann, 2001). A whole new socialization context emerges, with teachers as new authority figures and peers as new agents of socialization. This new context can be expected to exert influences on the political knowledge of children. Regarding gender differences, it may be important to note that the overwhelming majority of primary school teachers in Germany are female (in our sample, we had only one male teacher) (Stürzer, 2005, p. 47). A new female role model may especially influence girls (political) learning. Moreover, the general cognitive stimuli given may thwart a more traditional gender-specific family socialization so that girls political knowledge might benefit more than boys. Regarding the development of gender differences in political knowledge in primary school, we therefore expect that: HL2: The gender differences should be highest at the beginning of the study and lowest at the end of the study. Age Previous studies found that as children grow older and are more exposed to political information, their content of political knowledge increases (e.g. Allen, Kirasic, & Spilich, 1997, p. 170f; Garcia, 1973, p. 186; Jahoda, 1963, p. 146; Stevens, 1982, p. 167) and their knowledge becomes more precise (cf. Dowse & Hughes, 1971, pp ; Connell, 1971; Thompson, 1979, pp ). For instance, studies have shown that, with increasing age, children know more political actors and political parties (Hafner, 2006, p. 108; Connell, 1971). In line with developmental theory, age is, in this reasoning, used as a proxy for the general development of the skills and orientations of children (cf. Moore et al., 1985). H2a: Against this background, we can hypothesize, for our own study, that the older the kids are, the more they politically know. However, other studies with children indicate that the development of political knowledge is a gradual process in which the overall cognitive and intellectual development, and therefore school grades, represent a more important factor than simply the biological age (Gemmeke, 1998, p. 279; see for a similar reasoning Andrain, 1971). These findings indicate that the cognitive achievements of children could also be caused by the interaction between information available at different age levels and the relevant conceptual structures already possessed by children (Berti & Andriolo, 2001, p. 368). School forms encompass children with similar competency levels; age differences within one form mean that a cognitively fast six-year old can be compared to a slow seven-year old (Gemmeke, 1998, pp ). If children are much older than their classmates in primary school, we could assume that they are 5

7 slower in their cognitive development and were, therefore, either enrolled later in school or had to repeat a class. Therefore, we can formulate a competing hypothesis regarding age: H2b: Age should not have an effect on the levels of political knowledge of children at all, or older age should be correlated with lower levels of knowledge. Keeping this line of reasoning, we could expect the following for the longitudinal perspective: HL3a: The differences between the age groups remain consistent during primary school. Socioeconomic background The effects of socioeconomic factors seem to be particularly relevant for the political learning of young children (cf. Moore et al., 1985; especially Gemmeke 1998 for ethnic disadvantages) and have already been identified by early socialization studies (cf. Greenstein, 1965; Hess & Torney, 1967). The significance of socioeconomic factors for learning processes in general is, for instance, underlined by Lareau (2003, p. 239) who stresses the exceptional position of the middle class for orientations of children. Socioeconomic aspects of the children s immediate environment could affect their political orientations (cf. Garbarino, Kostelny, & Barry, 1997); the lower the socioeconomic status, the less active and involved in politics and other aspects of community life a child is likely to be (Greenstein, 1965, pp ). Studies investigating the political knowledge of children found similar relationships (cf. among others Dowse & Hughes, 1971, pp ; Easton & Dennis, 1969, p. 353; Lutkus & Weiss, 2007, p. 14; Orum et al., 1974, p. 204; Snyder, Dillow, & Hoffman, 2009, p. 189). Therefore we expect that: H3: Children with a better socioeconomic background have higher levels of political knowledge than those who are less well-off. School attendance provides cognitive input and enables contact with political issues for all children, regardless of their socioeconomic background. Therefore, the political knowledge of children with lower socioeconomic status should especially benefit from the time spent in school and be able to compensate for their lower knowledge levels at the time of their school enrollment. We argue here that children from a less privileged background now come into contact with political issues and can therefore catch up with their classmates. HL4: Regarding the longitudinal perspective, we would expect the differences to be highest at the beginning of school (wave 1). School attendance should then compensate at least partially for these differences so that they are lowest at the end of primary school (wave 3). 6

8 3.2 Migration US American political socialization studies have shown that children of different ethnic backgrounds also differ regarding their political involvement (Abramson, 1977; Garcia, 1973; Greenberg, 1970; Orum et al., 1974). Children of some ethnic minorities are less politically informed than others. More recent studies show similar results for the Netherlands and Germany (Gemmeke, 1994, 1998; van Deth, 2007; Vollmar, 2007; van Deth et al., 2011; Snyder et al., 2009, p. 191) and find that a migration background exerts a negative effect on cognitive political orientations. These lower political knowledge levels of children with migration backgrounds can partially be attributed to their otherwise lower socioeconomic status; both factors reinforce each other (Moore et al., 1985, pp ). 6 In Germany, a representative children s study shows that especially children from a Turkish family background also belong to the lowest social milieu (e.g. Betz, 2008, pp ). However, knowledge differences remain in many studies, even if controlled for other socioeconomic factors. H4a: We therefore expect that children with a migration background exhibit lower knowledge levels than their German classmates. In socialization and educational studies, it has additionally been found that the spoken language at home is a better indicator for analyzing the migration background of adolescents (see for Germany especially the results of the PISA study e.g Müller & Stanat, 2006; Walter, 2008). Niemi and Junn (1998) could, for instance, show that youth from families where other languages were spoken than English have less political knowledge than youth from families where English is the predominant language at home (Niemi & Junn, 1998, pp ). H4b: We therefore expect that children from families where the dominant language spoken at home is not German, display lower levels of political knowledge than children where German is the most spoken language in the family. Regarding the development of political knowledge of children with a migration background, we argue similar to the role of socioeconomic status above. Since school attendance provides cognitive input and enables contact with political issues for all children regardless of their background, it can be assumed that these children should especially benefit from primary school and may be able to compensate for their lower knowledge levels at the time of their school enrollment. HL5: Regarding the longitudinal perspective, we would expect the differences due to aspects of the family s migration background to be highest at the beginning of school (wave 1). Going to school should then partially compensate for these differences so that they are lowest at wave 3. 6 Findings of comparative educational studies, such as PISA, repeatedly show that, especially in Germany, children with an emigrational family background are disadvantaged regarding learning and school success in general (e.g. Auernheimer, 2006). Studies dealing with inequalities in youth political involvement also conclude that social and ethnic background are linked in complex ways (Verba, Burns, & Schlozman, 2003). 7

9 3.3 Political communication and media Media reception and communication confront individuals also with political phenomena. Early socialization studies have already pointed to the crucial impact of the mass media on children s perception of political life (cf. Atkin & Gantz, 1978; Conway, Wychoff, Feldbaum, & Ahern, 1981; Gemmeke, 1998; Rubin, 1978; Torney, Oppenheim, & Farnen, 1975; Connell, 1971). News broadcasting especially exerts a positive influence on political knowledge (cf. Bronstein et al., 1993; Chaffee, Jackson-Beeck, Durall, & Wilson, 1977; Conway, Stevens, & Smith, 1975; Conway et al., 1981). Media should not simply be regarded as a separate socialization influence, but rather as one which exists in relationship to the other main influences (family and school) at a young age (Samaniego & Pascual, 2007). For example, parents and teachers can influence socialization effects of the media on children, not only directly via deciding what the child is allowed to watch, but more importantly, by discussing media content with young children. Interpersonal communication, e.g. in form of political discussions with parents, teachers, and/or other children, is also important for civic socialization (McIntosh, Hart, & Youniss, 2007; Hafner, 2006, p. 125). It has been found that adolescents and children from families in which discussions about public matters are common and the reception of news broadcasting is frequent are likely to be better informed about politics than children from other families (e.g. National Center for Education Statistics, 1998, p. 10). Two hypotheses derive from this argument: H5: Children who watch news broadcasting are more politically knowledgeable than those who watch less or none. H6: Children who communicate with others about political events should be more knowledgeable than those who politically communicate less or not at all. Books are (still) important forms of media for children (e.g. Richter, 2014). Being read to by one s parents or reading books on your own is an important family activity in German families. Although its influence on political learning has not been studied so far, reading books (or being read to) has generally been found to have a positive influence on general learning processes (Klein & Biedinger, 2009). It is also a way to provide direct political information and stimulate discussions about political matters within the family or with peers. Therefore, we expect that: H7: Children who frequently read books are more politically knowledgeable than children who don t. HL7: Regarding the longitudinal perspective, we would expect that the differences occur in all three waves. 3.4 Cognitive competence and school motivation Studies with adults show (Delli Carpini & Keeter, 1996) that the overall cognitive competence of individuals is positively correlated with political knowledge. We can also assume this for children; 8

10 those who are competent and cognitively motivated in other domains could also benefit in their political learning from these cognitive skills. H8: Therefore, we expect that the more cognitively competent children are in general, the higher their political knowledge, as well. We can also assume that (political) learning processes are facilitated if children enjoy going to school; a good learning environment contributes to the acquisition of knowledge. Therefore we expect that; H9: Children who enjoy going to school are more politically knowledgeable than those who do not enjoy to do so. HL8: Regarding the longitudinal perspective, we would expect that these differences occur in all three waves. 4. Data and Methods The data of the following empirical investigation is drawn from the German study Demokratie Leben Lernen ( Learning to Live Democracy ). The study wants to improve the theoretical and empirical knowledge of political socialization processes in early childhood. Starting with the basic assumption that crucial impulses for the development of democratic personalities are already effective at a young age, the study seeks to obtain information on the political and social orientations of children and on the influence of the two main socialization contexts at a young age: family and school. A research design with three waves was set up first, at the beginning of the children s education, immediately after they start school; second, at the end of the first school year and; third, at the end of the fourth school year (see for more details van Deth et al., 2011; van Deth, Abendschön, Rathke, & Vollmar, 2007; Tausendpfund, 2008). The first methodological challenge was the development of a standardized questionnaire suitable for children who have not yet developed literacy (in the first school year). For this purpose, qualitative interviews with children and interviews with teachers and psychologists were conducted in a pilot phase of the project (Rathke, 2007). The resulting pictographic questionnaire for children covers about 100 items dealing with political knowledge, images of good citizenship, desirable social behavior, conceptions of democracy, political communication, gender roles and attitudes toward political issues and authorities. In a nutshell, the questionnaire in the first wave only consists of symbols and pictures, and instead of Yes, No or Don t Know, you ll find Smiley s with happy and sad faces and question marks. In addition to the questionnaire, a manual for the interviewers was developed specifying the exact wording of the questions which were to be read to the children in class. During a classroom session, one of the interviewers read each question aloud and showed an accompanying symbol with an overhead projector. Symbols are used in order to ensure that each child responds to each particular question read by the interviewer. For the third wave, the pictographic questionnaire was updated, meaning it was translated into a paper-pencil questionnaire (Tausendpfund, 2008) because the children were then able to read and write. After a short 9

11 introduction, the children completed the questionnaire independently. On the whole, 1070 children participated in one of the waves in the city of Mannheim (population 310,000). 7 In the first wave we have 736, in the second, 725, and in the third, 765 valid questionnaires. 431 children participated in all three waves. In our analysis, we use only the data of the boys and girls who participated in all three surveys. Operationalization of dependent variables As has been discussed previously in this paper, two of Neuman s (1986) components of political sophistication, namely political salience as the precondition of political sophistication and political knowledge as factual information about politics, are considered in our analysis. Four items deal with the perception and recognition of political actors and seem appropriate to capture the children s political salience (see Table 1). In the first and second survey wave, we only used dichotomous items with two response categories (Yes/No). In the third survey, two questions were more difficult; the children were asked for names of political parties (for example SPD or CDU) and the name of the German chancellor (Angela Merkel). Table 1: Question wording (political salience) Items First Wave Second Wave Third Wave Politicians Have you ever heard about politicians? Have you ever heard about politicians? Have you ever heard about politicians? Parties Have you ever heard about parties? Have you ever heard about parties? Do you know some political parties? Please list the names in the yellow field. Democracy Have you ever heard about Democracy? Have you ever heard about Democracy? Have you ever heard about Democracy? Chancellor Do you know this man? (picture of the former chancellor Gerhard Schröder) Do you know this man? (picture of the former chancellor Gerhard Schröder) If you know the name of the person, please write it in the yellow field (picture of the current chancellor Angela Merkel) Note: Questions wordings are translated from German (the original questionnaire is available on request). In the next step, we want to investigate whether a latent structure of political salience underlying the children s answers can be discovered. Principal component analyses with tetrachoric correlations could validate this question. As can be seen in Table 2, the dimensional analysis for all waves shows a clear one-dimensional solution which accounts for around 60 percent of the variance (a Mokkenanalysis leads to comparable results). We create a simple additive index of these items which can be labeled political salience (additive index ranging from 0 (no salience) to 1 (high salience)). 7 Mannheim is a very socially heterogeneous city in the German state of Baden-Württemberg. It was therefore decided to not further vary the local context in order to avoid additional complexity. The restriction on one city as a targeted population also has the advantage that all schools are subject to identical regulations and legal requirements and have the same financial budget per child. 10

12 Table 2: Factor Structures of Political Salience Factor loadings Items First Wave Second Wave Third Wave Politicians Parties Democracy Chancellor Eigenvalue Variance explained (%) Cronbach s Alpha N Note: Principal component analyses; Tetrachoric correlations; Listwise deletion of missing values. Political knowledge is measured with two separate items (see Table 3): one question inquires who has the most power in Germany ; another question asks what is democracy. Obviously, both questions are more difficult than the straightforward questions in Table 1. Each question has four possible responses, including the don t know category. Table 3: Question Wording (political knowledge) Items First Wave Second Wave Third Wave Power in Germany Who has the most power in Germany? Possible answers: 1) King, 2) Mayor, 3) Chancellor, 4) Don t know Who has the most power in Germany? Possible answers: 1) King, 2) Mayor, 3) Chancellor, 4) Don t know Who has the most power in Germany? Possible answers: 1) King, 2) Mayor, 3) Chancellor, 4) Don t know Meaning of Democracy What is democracy? Possible answers: 1) One citizen has the power, 2) All citizens have the power 3) Don t know What is democracy? Possible answers: 1) One citizen has the power, 2) All citizens have the power 3) Don t know What is democracy? Possible answers: 1) One citizen has the power, 2) All citizens have the power; 3) All citizens have money 4) Don t know Note: Questions wordings are translated from German. The variables were coded to differentiate between objectively correct and wrong (objectively wrong and don t know answers). The correlations between the two items are positive but very low (r<.10). Therefore, no index is created. Instead, we will consider the two items separately. Operationalization of independent variables The explanatory variables fall under five categories: socio-demographic characteristics, migration background, communication and media usage, as well as cognitive abilities. We start with gender and age as classical demographic characteristics. The variable gender is a dummy variable coded 0 for boy and 1 for girl. Although all children were in the same class in every survey wave, we note an age gap. The age difference between the youngest and oldest child is around two and a half years. For the analyses, the age was rescaled in every wave (youngest child=0; oldest child=1), so that the coefficients indicate the change associated with moving from the youngest to the oldest child. Unfortunately, we have no direct measurement of the socioeconomic situation of the 11

13 family or child. 8 Therefore, the socioeconomic status is constructed on the basis of socioeconomic characteristics of the child s neighborhood. In other words, we use the socioeconomic residence area as a proxy for the socioeconomic situation of the child. The information outlining the neighborhood s socioeconomic status includes its unemployment rate, its percentage of welfare recipients, and its ratio of migrants. The neighborhoods were ranked according to their values for each of these indicators. On the basis of these three characteristics, three groups were built: high, middle, and low. 9 To test the impact of migration background for political knowledge, we have several indicators. First of all, we distinguish between three ethnic groups: German, Turkish and other (very heterogeneous ethnic group). Secondly, we distinguish between two characteristics: child s citizenship (German, Turkish or other) and parents country of birth (four groups: 1. both were born in Germany, 2. both were born in Turkey, 3. both were born in other countries, 4. one parent (father or mother) was born in Germany). Six items are used to capture the children s political communication and media usage. Each child was asked if he/she discusses social and political themes with his/her parents, teachers, or other children (dummy variable coded 0 for No and 1 for Yes). In addition, the boys and girls were asked whether they watch children newscasts, adult newscasts and/or read books (also coded as dummy). Unfortunately, indicators for cognitive abilities are not available in the first two waves. In the third wave, we can, however, use a measurement of the academic self-concept from the Programme for International Student Assessment as an indicator for cognitive abilities. The measurement consists of three items and we created a simple additional index. In addition, we use the following indicator as a proxy for school motivation; I like to go to school. Multi-Level-Approach For the remainder of this article, our primary focus will be on testing the (individual) hypotheses presented previously. However, we need to consider the data structure. We have data from pupils in schools. Multi-level modeling is a particular, namely hierarchical, regression technique which is required if individual-level variables of a sample are nested within aggregate data as it is in our case: pupils in schools. This method simultaneously and statistically estimates accurately both the influence of contextual and individual-level factors (Snijders & Bosker, 1999). Admittedly, the focus of our analysis is, however, on the micro level In the parents survey, we asked for income, education and social status of the family. However, since the response rate of the survey was only at about 30 percent, we would lose too many cases for our three-wave-analysis here. Therefore, we will not use this information. 9 This means that, in a high neighborhood, the unemployment rate, the percentage of welfare recipients, and the ratio of migrants are low, whereas in a "low" neighborhood, the unemployment rate, the percentage of welfare recipients, and the ratio of migrants are high. 10 We have data from 18 schools and 34 classes. This means that the data structure correctly asks for a three level approach (pupils nested in classes nested in schools). However, due to low case number on pupils in classes and not to complicate interpretations, only a second level, the school level, is regarded here. The minimum of pupils in schools is 5, the average is 24, the maximum is

14 Share of the "Yes"-answers (Scale 0-1) 5. Empirical Analysis Descriptive findings Prior to the presentation of the multi-level models, we find it instructive to have a look at some descriptions which summarize the trends of political knowledge over the years. Figure 1 shows the development of political salience during the first four years in school. At first glance, we observe a considerable increase during the children s time in primary school. The index of political salience increases from 0.39 in the first wave to over 0.50 in the second wave to 0.69 in the last wave. T-tests reveal that the increase of the indices is also statistically significant. From first to second wave, the level of the index rises by.11 points ( t =6.6, df=430, p<0.00), and from the second to third wave by.19 points ( t =11.7, df=430, p<0.00). Figure 1: Trend of political salience (N=431) 1,0 0, ,8 0,7 0,6 0,5 0,4 0, ,2 0,1 0,0 wave1 wave2 wave3 Politicans Parties Democracy Chancellor Mean (Index) However, we observe considerable differences between the four items. The small and insignificant change from the second to the third wave of the Chancellor-Item can be highlighted in this context ( t =1.21, df=430, p>0.10). We assume that this is attributable to the indicator itself. In the first and second waves, the Chancellor item was only a question of recognizing the person from a picture or not (without having to know the exact name of the person); in the third wave, the children had not only to recognize the person, but also to know the name of the chancellor. It is obvious that the item in the third wave was more difficult than in the first or second wave, so we cannot (should not) rightfully compare the questions. Figure 2 provides a visual comparison of the change in the knowledge levels regarding the two items Meaning of Democracy and Power in Germany. It is particularly interesting to note that we cannot observe considerable differences in the Meaning of Democracy item in the first two waves. 13

15 Share of the "Correct"-Answers (Scale 0-1) We only see a statistically significant increase from wave two to wave three ( t =4.32, df=429, p<0.00). In addition, only one third of the children answered correctly. For most children, the question was even too difficult in fourth grade, as more than 60 percent answered with I don t know. Only a few children have chosen an explicitly wrong answer (less than four percent). The share of correct answers of the Power in Germany items increases from 30 percent in the first wave to 45 percent in the second wave and to 70 percent in the third wave. It is striking that the share of objectively wrong answers decreases over time from 59 percent in the first wave and 30 percent in the second wave to 20 percent in the third wave. The share of don t know answers descends from 20 percent (first/second wave) to nine percent (third wave). Figure 2: Trend of political knowledge (N=431) 1,0 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0, wave1 wave2 wave3 Meaning of Democracy Power in Germany The descriptive analyses in Figure 1 and Figure 2 suggest that the political knowledge level of children indeed rises from wave to wave. These findings provide clear evidence in favor of our hypothesis HL1. Multivariate findings In the next step, we want to take a more detailed look at the impact of our explanatory variables on political knowledge. We will compute several models. For a start, we consider the effects of sociodemographic characteristics, migration background, communication and media use, as well as cognitive abilities, on political knowledge separately. In a final step, we include all explanatory variables in a common model. To estimate the relationships between our explanatory variables and political knowledge, we use multilevel linear regression for political salience and multilevel logistic regression logistic regression models for the items Power in Germany and Meaning of Democracy. 11 To document the variation of the dependent variables at the individual and contextual levels, we begin by estimating an empty model. Building on the empty model, which excludes independent 11 The analysis is conducted using the statistical program Stata version 12.1 which offers various multi-level applications. 14

16 variables, we can establish the variance to be explained for the micro and macro levels (intraclass correlation). The results of these analyses are displayed in Table 4. The intraclass correlation coefficient (ICC) is defined as the variance of contextual level divided by the total variance. The ICC differs depending on the waves and the dependent variables. The ICC on political salience increases from wave to wave, whereas the ICC on Meaning of Democracy is negligible (the Likelihood-Ratio- Test is also insignificant). Individual level Table 4: Empty model Political Salience 1 Power in Germany 2 Meaning of Democracy 2 W1 W2 W3 W1 W2 W3 W1 W2 W3 Constant.39***.48***.66*** -.98*** -.33 #.78*** -1.12*** -1.14*** -.54*** Random effects Individual level Contextual level ICC.00.09***.19***.02.08**.08** N (schools) 431 (18) 431 (18) 431 (18) 431 (18) 415 (18) 431 (18) 421 (18) 430 (18) 431 (18) 1) Multi-level linear analysis with restricted-maximum-likelihood-estimation (unstandardized regression coefficients), 2) Multi-level logistic analysis with restrictedmaximum-likelihood-estimation (unstandardized logistic regression coefficients). Significance level: #=p<0.10; *=p<0.05; **=p<0.01; ***=p< Let us start with the effects of socio-demographic characteristics on political knowledge. The results of the regression analyses with gender, age and socioeconomic characteristics of the neighborhood are summarized in Table 5. Three findings are particularly remarkable. First of all, we see gender-based differences in political knowledge (H1a), but not in all knowledge domains (H1b). Boys score higher than girls on measures of political salience and Power in Germany, but not on Meaning of Democracy. Moreover, it is also striking that the effect of gender differences decreases over time. There are considerable differences between boys and girls in the first and second wave, but not in the third wave (HL2). Second, we see an unclear relationship between age and political knowledge. Depending on the wave and on the item, we find different results. On political salience (wave 3) and Power in Germany (wave 2), the impact is negative; on Meaning of Democracy, the effect is positive (wave 1). In this respect, we cannot confirm or reject our hypotheses H2a and H2b. We see no stable effect of age during primary school to a point that we can reject our hypothesis HL3. Third, the results strengthen our view that especially socioeconomic characteristics of the neighborhood are relevant to explain differences between the children. In neighborhoods with a low unemployment rate, a low rate of people receiving social benefits, and a low rate of migrants, political salience is higher than in a neighborhood with a high unemployment rate, a high rate of people receiving social benefits, and a high rate of migrants (H3). The same pattern can be seen regarding the Power in Germany item, but not regarding the Meaning of Democracy item. Contrary to our expectation, school attendance does not compensate differences between socioeconomic classes (HL4). On the contrary, the impact of socioeconomic characteristics on political salience increases over time. The share of variance explained rises from four to ten percent. All in all, these results suggest that there is a strong relationship between socio-demographic characteristics and political knowledge. Especially 15

17 socioeconomic factors have an impact on political knowledge, but we also see significant coefficients of gender and age. Table 5: Impact of socio-demographic factors on political knowledge (regression coefficients) Political Salience 1 Power in Germany 2 Meaning of Democracy 2 W1 W2 W3 W1 W2 W3 W1 W2 W3 Individual level Gender: girl -.12*** -.10** ** Age: standardized (0-1) * ** * SES residence area (Ref. middle) high.03.15**.10* # low -.08* * ** -.62*.45.62*.38 Constant.44***.47***.76*** **.73 # -2.21*** -1.47*** -.53 Random effects Individual level Contextual level ICC.00.04**.07** Explained variance N (schools) 431 (18) 431 (18) 431 (18) 431 (18) 415 (18) 431 (18) 421 (18) 430 (18) 431 (18) 1) Multi-level linear analysis with restricted-maximum-likelihood-estimation (unstandardized regression coefficients), 2) Multi-level logistic analysis with restrictedmaximum-likelihood-estimation (unstandardized logistic regression coefficients). 3) Multilevel linear regression: Snijders/Bosker R-squared (Level 1); Logistic Regression (Pseudo R²). Significance level: #=p<0.10; *=p<0.05; **=p<0.01; ***=p< Table 6 presents an overview of the impact of a migration background of the family on political knowledge. The table is, in several ways, quite revealing. First, the results of the two-level regression analysis show that children with a native German background are more likely to have high levels of political knowledge than children with an immigration background (H4a). As can be seen from the table particularly Turkish children scored significantly lower than the other two groups. All signs are negative and five of the nine coefficients are significant. Second, we also see an influence of the parental birthplace. Having non-german parents increases the child s probability to score low on political knowledge. Interestingly, this also holds true for children who only have one non-german parent. Third, the most striking result to emerge from the data in Table 6 is a language effect. It confirms our hypothesis H4b that not speaking German at home also increases the chance of lower levels of political knowledge compared to children in whose family the predominant language is German. Overall, these results indicate that a migration background of children s family background has a negative effect on political knowledge. The impact of migration factors on political knowledge also increases over time (with the exception of Meaning of Democracy ) and it is particularly pronounced regarding political salience (R² rises from four percent in the first wave, to 19 percent in the last wave). In other words, (primary) school cannot compensate the effects of migration on political knowledge (HL5). 16