Faculty of Social and Behavioural Sciences
Graduate School of Child Development and Education
Gender differences in citizenship
competences: The role of students’
perceived openness of classroom climate
Research Master Child Development and Education Research Master Thesis
L. Hoek, s11346515
Prof. dr. A. B. Dijkstra & R. Daas, MSc 10 July 2018
Abstract
Since 2005, Dutch schools are legally obliged to actively teach students knowledge, attitudes and skills that belong to citizenship. By doing so, schools help to sustain democratization and social cohesion in society. Primary analyses of the International Civic and Citizenship Education Study (ICCS) of 2016 indicate the presence of gender differences in citizenship competences. As previous research shows that girls generally perceive the classroom climate to be more open, the present study sought to understand the mechanism underlying these gender differences by examining the mediating role of students’ perceived openness of classroom climate. This was done using structural equation modelling while accounting for the nested structure of the data. Citizenship self-efficacy, societal participation, talking about political and societal matters and civic knowledge were set as outcome variables and provide a complete picture of students’ citizenship competences. Results show that the relationship between gender and citizenship self-efficacy, and gender and talking about political and societal matters, is fully mediated by open classroom climate. The relationship between gender and societal participation, and gender and civic knowledge, are partially mediated by open classroom climate. Future research can complement this study by including the aggregate of students’ perceived openness of classroom climate. In addition, observations could provide more information on the quality of classroom discussions.
Keywords: Citizenship education, Gender differences, Open classroom climate,
Citizenship competences, International Civic and Citizenship Education Study (ICCS) 2016.
Introduction
Dutch society and educational policy perceive citizenship education as an important task of education (Veugelers, 2007; Education Council, 2012; Inspectorate of Education, 2016). Due to the growing need for increasing social cohesion in society (Oser & Veugelers, 2008; Dijkstra, 2012) and as a result of increased (cultural) diversity in society, continuing individualization, growing polarization and hardening in the social climate (Education Council, 2003; Mattei & Broeks, 2018), the Dutch government decided in 2005 to legally oblige schools to actively teach students knowledge, attitudes and skills that belong to citizenship, which is defined as ‘the willingness and the ability to be part of a community and to actively contribute to this community’ (Ministry of Education, Culture and Science, 2005). By doing so, schools help to sustain democratization and social cohesion in society. Like the Netherlands, many countries throughout Europe and the rest of the world nowadays actively pay attention to citizenship education in school (Daas, 2014; Eurydice, 2017; Schulz et al., 2017).
The International Civic and Citizenship Education Study (ICCS) 2016 is a large-scale research project initiated by the International Association for the Evaluation of Educational Achievement (IEA) (Schulz et al., 2017). The study addresses students’ civic knowledge, understanding, perceptions, attitudes, engagement and behaviour with the aims of investigating ways in which students are prepared for their roles as citizens in the second decade of the 21st century. Citizenship competences, which can be described as the knowledge, skills, attitudes and reflection that students need in a democratic and multicultural society – with the aims of being able to fulfil social tasks that are part of daily life, differ over students of different ethnicities, socioeconomic backgrounds, education levels, and also between boys and girls (Geijsel et al., 2012). Primary analyses of the Dutch sample of the ICCS of 2016 (Munniksma et al., 2017) indicate that gender differences are visible on almost all components of citizenship competences. Interestingly, boys and girls in the Netherlands receive equal opportunities within the classroom to learn citizenship competences, yet boys generally obtain lower scores on citizenship competences compared to their female peers.
Schools play an important role in students’ effective learning of citizenship competences. Previous research shows that the quality of the pedagogical teacher-student relationship, providing students relevant learning situations, and a democratic and participatory school climate and classroom climate matter for students’ citizenship competences (Geboers, Geijsel, Admiraal, & Ten Dam, 2013; Isac, Maslowski, Creemers, & Van der Werf, 2013). An open and democratic classroom climate in which discussion and dialogue take place was for instance found to effectively promote the development of citizenship competences of secondary-school students (Geboers, Geijsel, Admiraal, & Ten Dam, 2013). In understanding the mechanism underlying gender differences in citizenship competences, students’ perceived openness of classroom climate is an important and influential school factor that might play an important role: is an open classroom climate equally beneficial for boys’ and girls’ acquaintance of citizenship competences?
An open classroom climate, which is defined as a classroom climate ‘that fosters discussion of controversial issues, encourages students to present diverse opinions, and emphasizes respect of those opinions from teachers and students’ (Godfrey & Grayman 2014, p. 1802), was found to be an important factor in influencing students’ expected political participation (Castillo, Miranda, Bonhomme, Cox, & Bascopé, 2015), and in fostering youth’s intention to be an informed voter (Campbell, 2008; Castillo et al., 2015). Moreover, gender differences in political participation tend to exist throughout adulthood (Coffé & Bolzendahl, 2010), which underscores the importance of providing students an open classroom climate that boys and girls benefit equally. In addition, exposure to an open classroom climate can partially compensate for the disadvantages in citizenship competences students coming from a low socioeconomic background experience (Campbell, 2008). An open classroom climate also has a positive and lasting influence on political trust (Claes, Hooghe, & Marien, 2012), which is a remarkable stable variable over short and long periods of time (Prior, 2010). Consequently, efforts to promote political interest in youth in school are strongly justified.
Previous research shows that an open classroom climate is positively related to several components of students’ citizenship competences, such as civic knowledge
(Campbell, 2007; Campbell, 2008; Alivernini & Manganelli, 2011; Gainous & Martens, 2013; Maurissen, Claes, & Barber, 2018; Knowles & McCafferty-Wright, 2015; Persson, 2015; Kuang, Kennedy, & Mok, 2018), students’ citizenship self-efficacy (Knowles & McCafferty-Wright, 2015; Manganelli, Lucidi, & Alivernini, 2015; Claes, Maurissen, & Havermans, 2017; Kuang, Kennedy, & Mok, 2018), talking about political and societal matters with family and friends (Maurissen, Claes, & Barber, 2018; Claes, Maurissen, & Havermans, 2017) and students’ societal participation (Claes, Maurissen, & Havermans, 2017). In other words: students who feel like they can speak up in the classroom and consequently perceive the classroom as more open might also be the students having more civic knowledge, feeling more confident about their own citizenship competences, participating more often in societal organizations, and talking more often about political and societal topics with family and friends.
Let us put this into an example. Suppose a teacher initiates a classroom discussion about refugees and migration. Students who perceive the classroom climate to be less open might engage less in this discussion, feel less encouraged to present diverse opinions, and feel less familiar with giving and receiving respect between teacher and (other) students in classroom discussions. This might translate into less developed citizenship competences. Students who are less engaged in classroom discussions, present few or none diverse opinions, and are less familiar with giving and receiving respect might know less of the different viewpoints that are presented in the classroom discussion and consequently gain less knowledge about refugees and migration. Little experience with participating in classroom discussions might also make these students feel less comfortable participating in discussions outside of the school context – for example in talking about political and societal matters with family and friends. Students might have a (controversial) opinion about a topic but feel like they cannot present it into the classroom as they perceive the classroom climate to be less open. Therefore, these students might not learn to argue and defend their opinion to other people and also might not learn to give and receive respect for different opinions – which could yield lower scores on citizenship self-efficacy. At last, hearing and understanding different viewpoints in classroom discussions and learning to respect people with other opinions could be motives for students to participate in society, for instance by doing volunteer
work with refugees as a result of a classroom discussion in which different viewpoints about the topic of refugees and migration were presented.
We now know that gender differences in citizenship competences exist (Munniksma et al., 2017) and that an open classroom climate positively influences students’ citizenship competences. As boys and girls may perceive the classroom climate differently, it is interesting to examine the role of students’ perceived openness of classroom climate in explaining gender differences in citizenship competences. Students’ perceived openness of classroom climate has, to my knowledge, not yet been investigated in studies examining gender differences in citizenship competences. Using data of ICCS 2016 (Schulz et al., 2017), the present study aims to contribute to this scientific niche by providing more insight in the underlying mechanism leading to the emergence of gender differences in citizenship competences. This is done by examining the mediating role of students’ perceived openness of classroom climate in the relationship between gender and citizenship competences.
Explaining gender differences in citizenship competences
Girls can generally adapt a little better to the school environment than boys, which can partially explain why girls outperform boys on academic outcomes (Spinath, Eckert, & Steinmayr, 2014). Boys also tend to show more noncompliance with school rules, such as not doing their homework or showing externalising behaviour, which may hinder them in their academic success. A study that followed American adolescents from grade five to grade eight found that boys were more likely to experience problematic behavioural and emotional school engagement, which was for instance visible in not attending school regularly or not caring much about teachers and school (Li & Lerner, 2011). The research finding of boys having lower levels of emotional school engagement than girls has also been confirmed in another study (Whitlock, 2006). In a study investigating the citizenship competences of Dutch adolescents, results demonstrate that girls generally developed more societal interest, prosocial ability and greater societal knowledge and interpersonal knowledge as compared to boys (Geboers, Geijsel, Admiraal, Jorgensen, & Ten Dam, 2015).
In practice this could mean that girls, who seem to be emotionally (e.g., caring about teachers and school) and behaviourally (e.g., attending school regularly) more engaged to what happens in school and in the classroom (Whitlock, 2006; Li & Lerner, 2011), and generally have better abilities to adapt to the school environment than boys (Spinath, Eckert, & Steinmayr, 2014), are more engaged to what happens in the classroom and consequently perceive the classroom climate to be more open. Moreover, girls’ societal interest, prosocial ability and societal and interpersonal knowledge (Geboers, Geijsel, Admiraal, Jorgensen, & Ten Dam, 2015) may also make it easier for girls to participate in classroom discussions – as they have greater interest in and knowledge of what happens in society (i.e., girls may be more informed and therefore easier form an opinion about a societal or political topic) and have better prosocial abilities and interpersonal knowledge which they can use in classroom discussions (i.e., it may be easier to give and receive respect for different opinions and girls may be more comfortable to respectfully react on the arguments posed by peers). In fact, several studies using European data of ICCS 1999 (Campbell, 2007; Barber, Sweetwood, & King, 2015) and ICCS 2009 (Claes, Maurissen, & Havermans, 2017; Maurissen, Claes, & Barber, 2018) found that girls are more likely to perceive the classroom climate as open compared to boys.
Thus, it may be the case that girls make better use of the provided opportunities to practice their discussion habits in classroom discussions, e.g., girls better learn ‘how to form and express an opinion and to listen respectfully to a counterpart’ (Maurissen, Claes, & Barber, 2018, p. 11) than boys, and consequently gain more knowledge, skills and understanding with respect to citizenship. The effect may be reinforcing: when becoming acquainted with discussions both inside and outside of the school context more regularly, students are more likely to frequently engage in discussions (Youniss, 2011). Moreover, the relationship between open classroom climate and citizenship competences may be reciprocal: Dutch girls generally have more knowledge, attitudes, skills and reflection with respect to citizenship than boys (Geijsel et al., 2012), which may make it easier for girls to participate in classroom discussions (i.e., with knowledge about a certain societal discussion topic, it is easier to participate in the discussion than without any knowledge on this societal discussion topic). As a consequence, girls may be more
likely to perceive the classroom as open and make better use of the provided opportunities to practice and gain knowledge, attitudes, skills and reflection with respect to citizenship.
Research Questions
In the present study, citizenship competences is operationalized as civic knowledge, citizenship self-efficacy, societal participation, and talking about political and societal matters with family and friends. The justification and measurement of these variables is discussed more elaborately in the methods section and in the report and assessment framework of ICCS 2016 (Schulz, Ainley, Fraillon, Losito, & Agrusti, 2016; Schulz et al., 2017). This study is a secondary analysis of the ICCS 2016 study. From the primary analyses of ICCS 2016 in the Dutch sample, we know that girls scored higher societal participation and civic knowledge, and that there was no effect of gender for citizenship self-efficacy and talking about political and societal matters with family and friends (Munniksma et al., 2017). The first sub-question is nonetheless included in the present paper to provide readers a complete picture.
The present study aims at investigating the role of students’ perceived openness of classroom climate in explaining gender differences in citizenship competences by addressing the following research question:
RQ Does students’ perceived openness of classroom climate mediate the relationship between gender and citizenship competences?
… and by addressing the following sub-questions:
SQ1 What is the relationship between gender and citizenship competences? SQ2 What is the relationship between gender and students’ perceived openness of
classroom climate?
SQ3 What is the relationship between students’ perceived openness of classroom climate and citizenship competences?
Hypotheses
Based on the primary analyses of the Dutch sample of ICCS 2016 (Munniksma et al., 2017), it is known that the relationship between gender and societal participation and gender and civic knowledge is positive. In other words: girls are expected to obtain higher scores on societal participation and civic knowledge. It is furthermore known that there is no effect of gender on citizenship self-efficacy and talking about political and societal matters with family and friends. The relationship between gender and students’ perceived openness of classroom climate is expected to be positive (Campbell, 2007; Barber, Sweetwood, & King, 2015; Claes, Maurissen, & Havermans, 2017; Maurissen, Claes, & Barber, 2018). Thus, girls are expected to perceive the classroom climate as more open as compared to boys.
Based on previous research, students’ perceived openness of classroom climate is expected to have a positive effect on citizenship self-efficacy (Knowles & McCafferty-Wright, 2015; Manganelli, Lucidi, & Alivernini, 2015; Claes, Maurissen, & Havermans, 2017; Kuang, Kennedy, & Mok, 2018), societal participation (Claes, Maurissen, & Havermans, 2017), talking about political and societal matters with family and friends (Maurissen, Claes, & Barber, 2018; Claes, Maurissen, & Havermans, 2017) and civic knowledge (Campbell, 2007; Campbell, 2008; Alivernini & Manganelli, 2011; Gainous & Martens, 2013; Maurissen, Claes, & Barber, 2018; Knowles & McCafferty-Wright, 2015; Persson, 2015; Kuang, Kennedy, & Mok, 2018). In other words: students who perceive the classroom climate to be more open are also expected to obtain higher scores on citizenship self-efficacy, societal participation, talking about political and societal matters with family and friends and civic knowledge.
With respect to the research question, it is therefore hypothesized that students’ perceived openness of classroom climate fully mediates the effect of gender on citizenship self-efficacy, and gender and talking about political and societal matters with family and friends, and partially mediates the effect of gender on societal participation, and on gender and civic knowledge.
Method Procedure
The ICCS is established by the IEA. The study addresses students’ civic knowledge, understanding, perceptions, attitudes, engagement and behaviour with the aims of investigating the ways in which students are prepared for their roles as citizens in the second decade of the 21st century. In ICCS 2016, data from 24 countries is gathered. The present study is based on the ICCS (Schulz et al., 2017) and used only data from the Dutch sample for secondary analyses (Munniksma et al., 2017).
Data Collection
The data collection took place between February and June 2016 (Schulz et al., 2017). The student civic and citizenship knowledge test consisted of 88 items measuring civic and citizenship knowledge, analysis, and reasoning. According to a balanced rotated design, the items were assigned to eight booklets, each of which contained three of a total eight eleven-item clusters. Each student completed one of the booklets. This took approximately 45 minutes. Along with the items, contextual material that served as a brief introduction to each item or set of items was provided. The student questionnaire lasted for approximately 30 to 40 minutes and was used to assess students’ perceptions about civics and citizenship. The student questionnaires also gained information on students’ backgrounds. An extensive overview of the data collection procedure can be found in the report and assessment framework of ICCS 2016 (Schulz et al., 2016; Munniksma et al., 2017; Schulz et al., 2017).
Population and Sample
The population can be defined as all students in grade 8 (i.e., students of approximately 14 years old). The ICCS 2016 (Schulz et al., 2017) makes use of stratified two-stage cluster samples. In the first stage, probabilities proportional to size (PPS) as measured by the number of students enrolled in a school were used to sample schools within the participating countries. The sampling team asked each participating country to recruit a minimum sample size of 150 schools. National characteristics were used to determine the
exact number of schools needed. In the second stage, representatives of the sampled participating schools provided a list of available grades 8. An intact class was randomly selected from this list. All students in that selected class were surveyed. Table 1 provides an overview of participants in the Netherlands. An extensive overview of the sampling procedure can be found in the report and assessment framework of ICCS 2016 (Schulz et al., 2016; Munniksma et al., 2017; Schulz et al., 2017).
Table 1
Overview of participants in the Netherlands
Total number of schools that participated in student survey
Total number of students assessed
Student participation rate (weighted) in %
123 2812 92.5
Source: Munniksma et al. (2017). Variables
With respect to citizenship competences, the present study focused on four scales of the ICCS as outcome variables: civic knowledge, citizenship self-efficacy, societal participation, and talking about political and societal matters with family and friends. These constructs of citizenship competences were selected because they tap on different aspects of citizenship competences (i.e., knowledge, skills, behaviour) and together provide a complete picture of students’ citizenship competences (i.e., in the present study, a student is considered to be competent in citizenship if he or she has civic knowledge, feels capable of one’s own citizenship competences, participates in society, and talks about political and societal matters with family and friends). As the four outcome variables are all constructs of students’ citizenship competences, they are supposed to have a mutual dependency. Therefore, they are modelled simultaneously as outcome variables and not separately. In what follows, the constructs that feature in the research question are more elaborately operationalized.
Civic knowledge. Civic knowledge entails knowledge on the way a democratic society ‘works’ (e.g., knowledge on laws, the parliament, or the way decisions are made) and knowledge of the underlying principles (e.g., equality of people, taking into account minority groups) and the application of these principles (e.g., elections and the right to vote) (Munniksma et al., 2017). The survey on students’ civic knowledge distinguishes four levels. Level D corresponds to students who have knowledge of basic concepts, like equality of people or the importance of secret voting for elections. Level C corresponds to students who can understand the important principles of a democratic society, like freedom or equality. Level B corresponds to students who understand how democracy works and who can, for instance, explain the function of the constitution. Level A corresponds to students who can relate the way societal and political processes work to its application and results. The variable is measured using 87 items. A more extensive overview of these levels can be found in the report of Munniksma et al. (2017) and in the report and assessment framework of ICCS 2016 (Schulz et al., 2016; Schulz et al., 2017). Citizenship self-efficacy. Students rated the extent to which they think they are capable of doing the following things: giving arguments for your opinion about a controversial political or societal matter; discussing a newspaper article about a conflict between countries; apply for candidate in a school election; bringing together a group of students to establish change in school; following a television debate about a controversial matter; writing a letter to the newspaper in which your opinion about a topical matter is expressed; speak to the class about a political or societal matter. Possible answers include: very well, reasonably well, not that well, not well at all. The reliability as measured by Cronbach’s alpha is 0.84 (nationally) and the variable is measured using 7 items (Munniksma et al., 2017).
Societal participation. Students indicated the extent to which they have been involved in activities of the following: a youth organization of a political party or trade union; a volunteer group that does things to help the neighbourhood; a group of youth that champions or campaigns for a certain matter. Possible answers include: in the past year, more than a year ago, never. The reliability as measured by Cronbach’s alpha is 0.63 (nationally) and the variable is measured using 7 items (Munniksma et al., 2017).
Talking about political and societal matters. Students were asked to indicate how often they do the following when they are not in school: talking about political or societal subjects with parents; talking about political or societal subjects with friends; talking about what happens in other countries with parents; talking about what happens in other countries with friends. Possible answers include: never or hardly ever, on a monthly base (i.e., at least once per month), on a weekly base (i.e., at least once per week), on a daily base, almost on a daily base. The reliability as measured by Cronbach’s alpha is 0.75 (nationally) and the variable is measured using 4 items (Munniksma et al., 2017).
Students’ perceived openness of classroom climate. Students rated the extent to which lessons in which political or societal subjects were discussed consisted of the following: teachers encourage students to form an opinion; teachers encourage students to proclaim their own opinion; students suggest recent political events to discuss in the classroom; students come out for their own opinion in the classroom, even though it differs from most students’ opinion; teachers encourage students to talk about subjects on which students have a different opinion; teachers expose different sides of a story when explaining a subject matter in class. Possible answers include: never, rarely, sometimes, and often. The reliability as measured by Cronbach’s alpha is 0.76 (nationally) and the variable is measured using 6 items (Munniksma et al., 2017).
Control variables. Three control variables are taken into account because they might be influential on the outcome variables and consequently cause biased research findings. The control variables of the present study include students’ socioeconomic status (SES), students’ expected education level and ethnicity. In this section, these three control variables will be discussed more elaborately.
Socioeconomic status. SES influences students’ citizenship self-efficacy,
attitudes and self-beliefs because it affects the opportunities and experiences of civic life and civic activities (Manganelli, Lucidi, & Alivernini, 2015). In addition, students coming from a low SES background are more likely to experience lower behavioural (e.g., attending school regularly) and emotional (e.g., caring about teachers and school) engagement within their schools (Li & Lerner, 2011). SES was found to have a small but positive partial effect on students’ citizenship knowledge and attitudes with respect to
citizenship (Geijsel et al., 2012). SES was also found to influence students’ perception of the classroom climate: students whose parents have a low education level (i.e., parents who did not finish secondary education) reported lower scores on the open classroom climate compared to students whose parents have a high education level (Claes, Maurissen, & Havermans, 2017). In general, SES has demonstrated to be very important to include in analyses on political attitudes and competences (Campbell, 2008; Castillo et al., 2015; Geijsel et al., 2012).
As 14-year-old students are likely to be unaware of the family income, ICCS 2016 used three proxies to measure students’ SES: parental occupation, the education level of parents, and the number of books in the house. Parental occupation and the number of books in the house are positively related to students’ civic knowledge: students with higher SES significantly scored better on civic knowledge than students with lower SES (Schulz et al., 2017). The education level of parents was too related to students’ civic knowledge: students of lower educated parents tend to have less civic knowledge compared to students with higher educated parents, and also to students’ citizenship efficacy: students of higher educated parents generally score higher on citizenship self-efficacy compared to students of lower educated parents (Munniksma et al., 2017).
Expected education level. The present study also included students’ expected
education level as a control variable using students’ self-reports. Students in a higher educational track appeared to have more civic knowledge compared to students in a lower educational track. Students’ expected education level did not influence students’ societal participation (Munniksma et al., 2017).
Ethnicity. Students’ ethnicity is included as a control variable using students’
self-reports. Research shows that classrooms that consist of students with different ethnicities generally have fewer discussions (Campbell, 2007). In a study examining citizenship competences of Dutch adolescents, students with parents having an immigrant background had a little less citizenship knowledge compared to students with parents without an immigrant background. However, with respect to attitudes, students’ citizenship self-efficacy and reflection within the domain of citizenship competences, students with parents having an immigrant background scored higher than students with
parents without an immigrant background (Geijsel et al., 2012). At last, a study found no significant effect of immigration background on students’ perceived openness of classroom climate (Claes, Maurissen, & Havermans, 2017), but this might be due to the fact that the variable of immigration background reflects whether students are first- or second-generation migrants, and not whether students belong to an ethnic minority group. In the present study, students with at least one parent born outside of the Netherlands were compared to students with both parents born in the Netherlands. Initial analyses with the ICCS 2016 data showed that students with a migration background generally have less civic knowledge but score higher on citizenship self-efficacy. Having a migration background did not influence societal participation (Munniksma et al., 2017). Data Analysis Plan
With the purpose of using structural equation model (SEM), a hypothesized path model consisting of gender, students’ perceived openness of classroom climate and four components of citizenship competences (i.e., civic knowledge, citizenship self-efficacy, societal participation, talking about political and societal matters with family and friends) was constructed (Figure 1). SEM is a set of techniques that explore the relationships among variables by analysing their covariance. SEM combines elements of traditional multivariate models (e.g., regression analysis, factor analysis, path analysis) and offers researchers many benefits over traditional multiple regression, such as the ability to account for less than perfect reliability for observed variables (i.e., the ability to account for measurement error) and the ability to model multiple outcomes simultaneously (Adelson, 2012). Analyses were conducted with R version 3.5.0 (R Core Team, 2018), using the lavaan package (Rosseel, 2012).
A subset of the data was created consisting of the variables that were used in the analysis of this study. The small amount of missing data of this subset (less than 1%) was handled using Missing Complete at Random (MCAR). The variables that consisted of ethnicity of the father and the mother were combined and dichotomized into either having both parents been born in the Netherlands (0) or either having at least one parent been born outside of the Netherlands (1). The variable expected education level was
dichotomized into lower expected education level (0) and higher expected education level (1). With respect to gender, boys were marked with 0 and girls were marked with 1.
As data are nested (students within schools), it was first assessed whether a multilevel SEM model needed to be applied. If variables at the between-level explain only a small percentage of variance relative to the total variance of the dependent variable, the modelling of between-level variables may not be meaningful with respect to the hypotheses. Prior to modelling, the influence of the dependency of the cluster was examined. In order to do so, a saturated model was fitted at the within level and a null model at the between level to test the significance of level-2-variance. This yielded a significant Chi-square (χ2 (6) = 202.499, N = 2711, p <.001), indicating that the
between-level variance is significantly larger than zero. In addition, the intra-class correlations (ICCs) were checked. Except for civic knowledge (ICC = 0.564), the ICCs were generally low (Table 2), indicating that there is low dependency to the cluster (i.e., only a small amount of variance in the dependent variables could be explained by its cluster). In addition, given that the research questions of the present study only involve variables at the student level, two-level analyses were considered not meaningful and the data analyses proceeded with single-level analyses. I accounted for possible between-group similarities by correcting standard errors and fit measures by using the lavaan.survey package (Oberski, 2014).
In accordance with the assessment framework (Schulz et al., 2016), sampling weights were added to do justice to the disproportional selection probabilities of the schools, the students, and the teachers (i.e., if a unit had a small selection probability, a large weight was added to compensate for this and vice versa). The sampling weights were also used to compensate for sample size loss (i.e., the sampling weights were multiplied by non-response adjustments). The total weight that is used in the analyses is a product of weight factors and adjustment factors and therefore reflects the selection probabilities and the non-response patterns. As the present study used single-level student-level analyses, the total student weight (TOTWGTS) was applied. In R, the survey package (Lumley, 2004; Lumley, 2017) was used to convert the data, as well as the cluster (school) and the total student weight, into a survey-design-object. The
unweighted model fit was converted into another object using the lavaan package (Rosseel, 2012). Both objects were passed into the lavaan.survey function (Obserski, 2014) to calculate the weighted model fit.
Table 2
Weighted correlations, intra-class correlations and descriptive statistics (mean and standard deviation)
1. 2. 3. 4. 5. 6. 7. 8. 9. Mean SD Min Max %
missing 1. Gender 1.000 0.511 0.500 0.00 1.00 0.000% 2. Open. clim. 0.125 1.000 47.458 8.558 16.67 78.16 0.011% 3. Cit. eff. -0.011 0.242 1.000 48.045 9.187 16.59 78.42 0.015% 4. Soc. part. 0.141 0.205 0.167 1.000 48.276 7.909 38.11 86.22 0.006% 5. Pol. disc. 0.037 0.282 0.303 0.247 1.000 49.872 8.485 34.77 80.01 0.006% 6. Know. 0.081 0.222 0.023 0.013 0.249 1.000 531.620 88.973 239.87 750.26 0.000% 7. SES -0.018 0.097 0.115 0.066 0.249 0.433 1.000 195.940 89.712 1.00 345.00 0.014% 8. Exp. ed. -0.037 0.106 0.116 0.038 0.220 0.378 0.299 1.000 0.349 0.477 0.00 1.00 0.009% 9. Ethn. -0.010 0.008 0.005 0.011 0.064 -0.009 0.025 0.022 1.000 0.106 0.308 0.00 1.00 0.016% ICC 0.018 0.075 0.037 0.020 0.093 0.564 0.273 0.232 0.041 - -
Note. Open. clim. = students’ perceived openness of classroom climate; cit. eff. = citizenship self-efficacy; soc. part. = societal participation; pol. disc. = talking about political and societal matters with family and friends; know. = civic knowledge; SES = students’ socio-economic status; ethn = ethnicity; ICC = intra-class correlation; SD = standard deviation.
Model Chi-square test. In this study, the steps as recommended by Kline (2010) were followed. First, the model was specified, and the fit measures were selected. Before interpreting the model estimates, it was assessed whether the model fit was adequate. The model fit indicates how well the model can reproduce the correlations between variables in the observed data and can be checked by looking at model fit indexes. A test to assess the model fit is the model Chi-square. In this study, the Chi-square test statistic was used to examine whether the model-implied covariance matrix is consistent with the data-implied covariance matrix (containing the interrelations that are observed in the data). The null hypothesis implies that both matrices are consistent and failure to reject the null hypothesis can be interpreted as consistent model fit. When using the Chi-square test to assess model fit, one needs to be careful if the sample size is large and if variables have large correlations, because this may yield unreliable model estimates. In the present study the correlations among the variables were moderate (ranging from 0.433 to -0.037) (see Table 2) and therefore appropriate for using the Chi-square test to assess model fit. However, the sample size for the present study is 2711. This is likely to influence the model Chi-square test. Therefore, other fit measures were additionally used to assess model fit.
Comparative fit index. The comparative fit index (CFI) provides an indication of how well the model fit improved compared to the null model, which generally assumes no covariance between observed variables. The CFI also considers the complexity of the model, and ranges from 0 to 1. The CFI is suitable when the number of degrees of freedom is small. Therefore, this study assessed the CFI as an additional measure of model fit. A CFI above 0.95 is preferred (Hu & Bentler, 1999) and the current study will follow this preference.
Root mean square error of approximation. Whereas the CFI is considered a “goodness-of-fit” index, one may also look at the “badness-of-fit” (Adelson, 2012). This can be done by examining the root mean square error of approximation (RMSEA). The
RMSEA takes into account the model complexity by estimating the amount of error of
approximation per model degree of freedom. The RMSEA also takes into account sample size. Greater values of the RMSEA indicate worse model fit. Rules of thumb of the
RMSEA include: an RMSEA lower than 0.05 indicates ‘close fit’; an RMSEA between
0.05 and 0.08 indicates a ‘a reasonable error of approximation’ or satisfactory fit, and an
RMSEA above 0.10 indicates ‘poor fit’ (Browne & Cudeck, 1992). An RMSEA between
0.08 and 0.10 was suggested to indicate ‘mediocre fit’ (MacCallum, Browne, & Sugawara, 1996). These rules of thumb for the RMSEA are followed in the present study. Moreover, the present study uses the accompanying 90% confidence intervals of the
RMSEA to assess model fit. Ideally, the lower bound of the 90% confidence interval
includes or is very near zero (and not larger than 0.05) and the upper bound is not very large, i.e., less than 0.08. The width of the confidence interval provides information about the precision in the estimate of the RMSEA. The present study used multiple model fit indexes (i.e., Chi-square test, CFI and RMSEA with confidence intervals) to assess model fit. This provides a complete picture of model fit.
Correlation residuals. When examining possible model improvements, the correlation residuals were consulted. Correlation residuals larger than 0.10 indicate that the relation between these two variables is not well explained by the model. After examining the correlation residuals, it was considered whether to add a direct effect between these variables, or to add a covariance between the residuals of the variables. The latter was applied if a direct effect between two variables did not seem in accordance with the theory.
Model selection procedure. An overview of the model selection procedure can be found in Table 3. First, a partial-mediation path analysis (Figure 1) was performed in R. This initial analysis resulted in a poor model fit (χ2 (6) = 202.499, N = 2711, p <.001), CFI =
0.832, RMSEA = 0.110 with a 90% confidence interval from 0.100 to 0.120). According to the Chi-square test, exact fit was rejected (i.e., the model has no exact fit). The RMSEA indicates poor model fit, and according to the accompanying confidence interval neither poor approximate fit or satisfactory approximate fit could be rejected. The CFI indicates poor approximate fit relative to the independence model. The p-value is significant, indicating that there is a significant deviation between the model-implied covariance matrix and the observed covariance matrix, and that the model does not fit the data well.
Figure 1. Partial-mediation model.
Second, it was assessed whether the direct effects of gender on citizenship self-efficacy, societal participation, talking about political and societal matters with family and friends, and civic knowledge were partially or fully mediated by students’ perceived openness of classroom climate. In model 1a, the direct effect of gender on citizenship self-efficacy was left out of the model in order to assess whether the model with partial mediation (model 1) or full mediation (model 1a) for this effect yielded a better fit. According to the Chi-square difference test, model 1a did not fit significantly worse relative to model 1 (p = .328). The effect of gender on citizenship self-efficacy thus seemed to be fully mediated by students’ perceived openness of classroom climate. In model 1b, the direct effect of gender on societal participation was left out of the model in order to assess whether the model with partial mediation (model 1) or full mediation (model 1b) for this effect yielded a better fit. According to the Chi-square difference test, model 1b fitted significantly worse relative to model 1 (p ≤ 0.001). The effect of gender on societal participation thus seemed to be partially mediated by students’ perceived openness of classroom climate. In model 1c, the direct effect of gender on talking about political and societal matters with family and friends was left out of the model in order to assess whether the model with partial mediation (model 1) or full mediation (model 1c) for this effect yielded a better fit. According to the Chi-square difference test, model 1c
did not fit significantly worse relative to model 1 (p = .207). The effect of gender on talking about political and societal matters with family and friends thus seemed to be fully mediated by students’ perceived openness of classroom climate. At last, in model 1d, the direct effect of gender on civic knowledge was left out of the model in order to assess whether the model with partial mediation (model 1) or full mediation (model 1d) for this effect yielded a better fit. According to the Chi-square difference test, model 1d fitted significantly worse relative to model 1 (p = .011). The effect of gender on civic knowledge thus seemed to be partially mediated by students’ perceived openness of classroom climate.
In model 2, the direct effects of gender on both citizenship self-efficacy and talking about political and societal matters with family and friends were left out of the model since the previous model selection step indicated that these effects were fully mediated by students’ perceived openness of classroom climate (Figure 2). This seemed to be in line with primary analyses of the ICCS 2016 data of the Netherlands, in which no significant gender differences were found for citizenship self-efficacy and talking about political and societal matters with family and friends (Munniksma et al., 2017). Model 2 still resulted in a poor model fit (χ2 (8) = 210.985, N = 2711, p <.001), CFI = 0.826,
RMSEA = 0.097 with a 90% confidence interval from 0.088 to 0.106). According to the
Chi-square test, exact fit was rejected (i.e., the model has no exact fit). The RMSEA indicates poor model fit, and according to the accompanying confidence interval neither poor approximate fit or satisfactory approximate fit could be rejected. The CFI indicates poor approximate fit relative to the independence model. The p-value is significant, indicating that there is a significant deviation between the model-implied covariance matrix and the observed covariance matrix, and that the model does not fit the data well. According to the Chi-square difference test, model 2 does not fit significantly worse relative to model 1 (p = .283).
Figure 2. Adapted model.
Third, the correlation residuals were assessed. The correlations between all four outcome variables were larger than 0.10. This indicates that the following relations are not well explained by the model: students’ citizenship self-efficacy and students’ societal participation; students’ citizenship self-efficacy and the extent to which students talk about political and societal matters with family and friends; students’ citizenship self-efficacy and students’ civic knowledge; students’ societal participation and the extent to which students talk about political and societal matters with family and friends; students’ societal participation and students’ civic knowledge; and at last the extent to which students talk about political and societal matters with family and friends and students’ civic knowledge. Adding these correlation residuals to the model yielded model 3 (Figure 3).
Figure 3. Final model.
Model 3 indicated a close model fit (χ2 (2) = 3.158, N = 2711, p = .206), CFI =
0.999, RMSEA = 0.015 with a 90% confidence interval from 0.000 to 0.038). According to the Chi-square test, exact fit could not be rejected. The RMSEA indicates close model fit, and the upper bound of the accompanying confidence interval is below .05, indicating that unsatisfactory fit could be could be rejected. The lower bound of the RMSEA confidence interval includes zero, indicating that close fit could not be rejected. The CFI is above 0.95 and therefore indicates good approximate fit relative to the independence model (Hu & Bentler, 1999). The p-value is not significant, indicating that there is no significant deviation between the model-implied covariance matrix and the observed covariance matrix, and that the model fits the data well. According to the Chi-square difference test, model 2 fits significantly worse relative to model 3 (p ≤ 0.001).
Table 3
Model selection procedure
χ2 Df P-value CFI AIC RMSEA RMSEA CI
Model 1 202.499 6 .000*** 0.832 148522.639 0.110 [0.100;0.120] Model 1a 205.742 7 .000*** 0.830 148522.176 0.102 [0.093;0.112] Anova (1 – 1a) 0.328 Model 1b 237.292 7 .000*** 0.803 148567.507 0.110 [0.101;0.120] Anova (1 – 1b) ≤ 0.001*** Model 1c 208.519 7 .000*** 0.828 148522.980 0.103 [0.094;0.112] Anova (1 – 1c) .207 Model 1d 207.800 7 .000*** 0.828 148532.374 0.103 [0.094;0.112] Anova (1 – 1d) .011* Model 2 210.985 8 .000*** 0.826 148522.517 0.097 [0.088;0.106] Anova (1 – 2) .283 Model 3 3.158 2 .206 0.999 148182.650 0.015 [0.000;0.038]
Anova (3 – 2) ≤
0.001***
Note. * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001, df = degrees of freedom, CI = confidence interval. Model 1 = partial-mediation model; model 1a = partial-mediation model without direct effect of gender on citizenship self-efficacy; model 1b = partial-mediation model without direct effect of gender on societal participation; model 1c = partial-mediation model without direct effect of gender on talking about political and societal matters with family and friends; model 1d = partial-mediation model without direct effect of gender on civic knowledge; model 2 = partial-mediation model without direct effect of gender on citizenship self-efficacy and talking about political and societal matters with family and friends; model 3 = partial-mediation model without direct effect of gender on citizenship self-efficacy and talking about political and societal matters with family and friends with correlation residuals between outcome variables.
Results
The results of the analyses can be found in Figure 4 and Figure 5 and more elaborately in Table 5. The effects of the control variables can be found in Table D in the Appendix.1 Overall, the effects of SES and expected education level on societal participation were not significant; nor were the effects of ethnicity on citizenship self-efficacy, societal participation, civic knowledge, and on students’ perceived openness of classroom climate. Figure 4 (direct effects) and Figure 5 (indirect and total effects) show standardized effect sizes to enable making meaningful comparisons between effects. The standardized effects can be interpreted as units of change in the standard deviations of the dependent variables with one standard deviation change in the independent variable (i.e., being a girl). In this section, I first discuss the explained variance of the final model. Second, I discuss the results with respect to the research questions. The effects of model 1 and model 2 can be found in Table B and Table C in the Appendix.
1 In the present study, the effects of the control variables were only regressed on the mediating variable
(open classroom climate) and on the outcome variables (citizenship self-efficacy, societal participation, talking about political and societal matters with family and friends, civic knowledge). However, one could also argue that students’ expected education level and students’ ethnicity have an influence on gender: students with a higher expected education level could more likely be girls, and students with a migration background could more often be boys. Therefore, in retrospective, students’ expected education level and students’ ethnicity were also regressed on gender. These effects however appeared non-significant and were not further included in the model.
Figure 4. Direct effects final model.
Figure 5. Indirect (and total) effects final model.
Explained variance. The explained variance did not differ largely throughout the model selection procedure (see Table A in Appendix). The analysis of the final model showed
that 3% of the variance in students’ perceived openness of classroom climate is explained by the model. For students’ citizenship self-efficacy, 7.6% of the variance was explained by the model. For students’ societal participation, 5.8% of the variance was explained by the model. For students’ talking about political and societal matters with family and friends, 15.6% of the variance was explained by the model. At last, for civic knowledge 30.6% of the variance was explained by the model.
Gender and citizenship competences (SQ1). The direct effect of gender on students’ societal participation is positive and small: being a girl resulted in 0.126 SDs increase students’ societal participation (p ≤ 0.001, CI = 1.316;2.701). This is in line with the hypotheses: if a student is a girl, she is more likely to participate in society compared to boys. The direct effect of gender on students’ civic knowledge is positive and small: being a girl resulted in 0.050 SDs increase students’ civic knowledge (p = 0.018, CI = 1.594;16.939). This is too in line with the hypotheses: if a student is a girl, she is likely to have more civic knowledge compared to male peers. The direct effects of gender on citizenship self-efficacy and talking about political and societal matters with family and friends were not significant (see Table B in Appendix) and were left out of the model in order to improve the model fit.
Gender and perceived classroom climate (SQ2). The direct effect of gender on students’ perceived openness of classroom climate, given the other variables in the model, is positive and small: being a girl resulted in 0.123 SDs increase students’ perceived openness of classroom climate (p ≤ 0.001, CI = 1.347;2.842). This is in line with the hypotheses: if a student is a girl, she is likely to perceive the classroom climate as more open compared to boys.
Perceived classroom climate and citizenship competences (SQ3). The direct effect of students’ perceived openness of classroom climate on students’ civic skills is positive and small: one SD increase in students’ perceived openness of classroom climate results in 0.250 SDs increase in students’ civic skills (p ≤ 0.001, CI = 0.222;0.322). The direct effect of students’ perceived openness of classroom climate on students’ societal participation is positive and small: one SD increase in students’ perceived openness of classroom climate results in 0.181 SDs increase in students’ societal participation (p ≤
0.001, CI = 0.123;0.213). The direct effect of students’ perceived openness of classroom climate on students’ talking about political and societal matters with family and friends is positive and small: one SD increase in students’ perceived openness of classroom climate results in 0.269 SDs increase in students’ talking about political and societal matters with family and friends (p ≤ 0.001, CI = 0.221;0.319). The direct effect of students’ perceived openness of classroom climate on students’ civic knowledge is positive and small: one
SD increase in students’ perceived openness of classroom climate results in 0.155 SDs
increase in students’ civic knowledge (p ≤ 0.001, CI = 1.092;2.253).
Gender and citizenship efficacy (RQa). The effect of gender on citizenship self-efficacy is fully mediated by students’ perceived openness of classroom climate. The indirect effect of gender on students’ citizenship self-efficacy via students’ perceived openness of classroom climate is positive and small: being a girl resulted in 0.031 SDs increase in students’ citizenship self-efficacy via students’ perceived openness of classroom climate (p ≤ 0.001, CI = 0.355;0.785). If a student is a girl, she is more likely to perceive the classroom climate as open and consequently obtain higher scores on citizenship self-efficacy compared to boys.
Gender and societal participation (RQb). The effect of gender on students’ societal participation is partially mediated by students’ perceived openness of classroom climate. The indirect effect of gender on students’ societal participation via students’ perceived openness of classroom climate is positive and small: being a girl resulted in 0.022 SDs increase in students’ societal participation via students’ perceived openness of classroom climate (p ≤ 0.001, CI = 0.192;0.513). If a student is a girl, she is more likely to perceive the classroom climate as open and consequently, she is more likely to participate in society compared to boys. The total effect of gender on students’ societal participation is positive and small: being a girl resulted in 0.149 SDs increase in students’ societal participation both directly and indirectly via students’ perceived openness of classroom climate (p ≤ 0.001, CI = 1.653;3.069). The direct effect of gender on students’ societal participation is larger than the indirect effect via students’ perceived openness of classroom climate and is therefore likely to account for most of the total effect.
Gender and talking about political and societal matters (RQc). The effect of gender on talking about political and societal matters with family and friends is fully mediated by students’ perceived openness of classroom climate. The indirect effect of gender on students’ talking about political and societal matters with family and friends via students’ perceived openness of classroom climate is positive and small: being a girl resulted in 0.033 SDs increase in students’ talking about political and societal matters with family and friends via students’ perceived openness of classroom climate (p ≤ 0.001, CI = 0.325;0.805). If a student is a girl, she is more likely to perceive the classroom climate as open and consequently, she is more likely to talk about political and societal matters with family and friends compared to boys.
Gender and civic knowledge (RQd). The effect of gender on civic knowledge is partially mediated by students’ perceived openness of classroom climate. The indirect effect of gender on students’ civic knowledge via students’ perceived openness of classroom climate is positive and small: being a girl resulted in 0.019 SDs increase in students’ civic knowledge via students’ perceived openness of classroom climate (p ≤ 0.001, CI = 1.845;5.162). If a student is a girl, she is more likely to perceive the classroom climate as open and consequently, she is more likely to gain more civic knowledge compared to boys. The total effect of gender on students’ civic knowledge is positive and small: being a girl resulted in 0.069 SDs increase in students’ civic knowledge both directly and indirectly via students’ perceived openness of classroom climate (p = 0.001, CI = 5.077;20.464). The direct effect of gender on students’ civic knowledge is larger than the indirect effect via students’ perceived openness of classroom climate and is therefore likely to account for most of the total effect.
Table 5
Direct and indirect effects of gender and open classroom climate
Exogenous variables
Gender Open classroom climate
Endogenous variables
Unst. [95% CI] Std. Unst. [95% CI] Std.
Citizenship self-efficacy Direct - - - 0.272 [0.222;0.322] 0.250*** Indirect 0.570 [0.355;0.785] 0.031*** - - - Total 0.570 [0.355;0.785] 0.031*** 0.272 [0.222;0.322] 0.250*** Societal participation Direct 2.009 [1.316;2.701] 0.126*** 0.168 [0.123;0.213] 0.181*** Indirect 0.352 [0.192;0.513] 0.022*** - - - Total 2.361 [1.653;3.069] 0.149*** 0.168 [0.123;0.213] 0.181***
Talking about political and societal matters
Direct - - - 0.270 [0.221;0.319] 0.269*** Indirect 0.565 [0.325;0.805] 0.033*** - - - Total 0.565 [0.325;0.805] 0.033*** 0.270 [0.221;0.319] 0.269*** Civic knowledge Direct 9.267 [1.594;16.939] 0.050* 1.673 [1.092;2.253] 0.155*** Indirect 3.504 [1.845;5.162] 0.019*** - - - Total 12.770 [5.077;20.464] 0.069*** 1.673 [1.092;2.253] 0.155*** Open classroom climate 2.094 [1.347;2.842] 0.123*** - - -
Note. * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001, unst = unstandardized, CI = confidence interval, std = standardized
Discussion
The present study aimed to contribute to a better understanding of the mechanism underlying gender differences in citizenship competences. This was done by focusing on the possible mediating role of students’ perceived openness of classroom climate. The present study is, to my knowledge, the first to use students’ perceived openness of classroom climate in trying to understand the underlying mechanism leading to gender differences in citizenship competences. It therefore delivers an important contribution to this scientific niche.
The research findings demonstrate that girls generally obtained significant higher scores on societal participation and on civic knowledge than boys, given the other variables in the model. The direct effects of gender on citizenship self-efficacy and talking about political and societal matters with family and friends were not significant and were left out of the final model: boys and girls did not score significantly different on these components of citizenship competences. The relations between gender and citizenship self-efficacy and gender and talking about political and societal matters with family and friends were only allowed to exist indirectly via students’ perceived openness of classroom climate. This is in line with what is known from the primary analyses of the Dutch sample of ICCS 2016 (Munniksma et al., 2017).
The research findings also demonstrate a significant effect of gender on students’ perceived openness of classroom climate, given the other variables in the model: girls generally perceive the classroom climate to be more open as compared to boys. This is in line with the expectations (Whitlock, 2006; Li & Lerner, 2011; Spinath, Eckert, & Steinmayr, 2014; Geboers, Geijsel, Admiraal, Jorgensen, & Ten Dam, 2015) and with previous research (Campbell 2007; Barber, Sweetwood, & King, 2015; Claes, Maurissen, & Havermans, 2017; Maurissen, Claes, & Barber, 2018). It nevertheless poses an interesting question: why do girls perceive the classroom climate to be more open as compared to their male peers? Additional qualitative research, for example using in-depth interviews, can provide more insight in why boys and girls perceive the classroom climate like they do.
An open and democratic classroom climate in which dialogue and discussions occur was demonstrated to be beneficial for students’ citizenship competences, which is in line with previous research (Geboers, Geijsel, Admiraal, & Ten Dam, 2013). More specific, the results of the present study point out that students’ perceived openness of classroom climate positively influences citizenship self-efficacy, societal participation, talking about political and societal matters with family and friends, and civic knowledge. This is in line with the hypotheses based on previous research demonstrating the positive relationship between an open classroom climate and citizenship self-efficacy (Knowles & McCafferty-Wright, 2015; Manganelli, Lucidi, & Alivernini, 2015; Claes, Maurissen, & Havermans, 2017; Kuang, Kennedy, & Mok, 2018), societal participation (Claes, Maurissen, & Havermans, 2017), talking about political and societal matters with family and friends (Maurissen, Claes, & Barber, 2015; Claes, Maurissen, & Havermans, 2017), and civic knowledge (Campbell, 2007; Campbell 2008; Alivernini & Manganelli, 2011; Gainous & Martens, 2013; Maurissen, Claes, & Barber, 2018; Knowles & McCafferty-Wright, 2015; Persson, 2015; Kuang, Kennedy, & Mok, 2018).
At last, the research findings with respect to the main research question demonstrate that the relationship between gender and citizenship self-efficacy, and gender and talking about political and societal matters, is fully mediated by students’ perceived openness of classroom climate and that the relationship between gender and societal participation, and gender and civic knowledge, is partially mediated by students’ perceived openness of classroom climate. These effects were demonstrated while controlling for students’ SES, ethnicity and expected education level. This research finding underscores the important role of students’ perceived openness of classroom climate in students’ citizenship competences. It shows that girls perceive the classroom climate to be more open as compared to boys, and that girls consequently obtain higher scores on citizenship self-efficacy, societal participation, talking about political and societal matters with family and friends, and on civic knowledge. The direct effects (i.e., gender on open classroom climate, and open classroom climate on all four constructs of citizenship competences) were remarkably larger than the indirect effects (i.e., gender via open classroom climate on all four constructs of citizenship competences). It indicates that open classroom climate does mediate the effect between gender and students’
citizenship competences, but the effects are very small. The direct effect of gender on civic knowledge was also remarkably smaller (0.050) than the effect of gender on societal participation (0.126). Girls do have more civic knowledge than their male peers, but the effect is very small.
An important limitation of the present study is the measurement of openness of classroom climate. The construct as used in the present study measures the extent to which students perceive the classroom climate as one in which they can exchange views and ideas of political and/or social topics. As Maurissen, Claes and Barber (2018) have also pointed out, this measurement does not include information about the quality of discussions, nor about how frequent classroom discussions occur. Future research using interviews and/or observations complemented by text analysis could provide additional information about what is being said by who in classroom discussions. Schuitema, Radstake, Van de Pol and Veugelers (2017) for instance found that when teachers strongly guide classroom discussions, it positively influenced the high content quality of classroom discussions, as well as student participation in classroom discussions. By guiding a classroom discussion, teachers can introduce other perspectives on a subject matter or ask students critical questions to think about. This can improve the quality of the content of classroom discussions and more students feel like they can participate in the classroom discussion – and thus perceive the classroom climate to be open.
Another limitation of the measurement of students’ perceived openness of classroom climate relates to the fact that it is an individual perception of the classroom climate. One could argue that the aggregate of perceived openness of classroom climate also needed to be included in the present study with the aims of obtaining a more general view of what the classroom climate looked like. Persson and Svensson (2017) for instance found that a class with a higher proportion of girls was associated with a lower likelihood of perceiving the classroom climate as messy and disorderly – the same might be the case for perceiving the classroom climate as open. However, the present study values precisely the individual perception of the classroom climate: being a boy or a girl influences a students’ individual perception of the classroom climate and this influences individual scores on citizenship self-efficacy, societal participation, talking about political
