Intergenerational transmission and the reinforcement of the political engagement gap: Identifying how university-educated parents enable their children to become more politically interested during early adolescence

Research consistently shows that parents' educational attainment is associated with their children's level of political interest. The life stage when this relationship is established and grows has been identified to be between the ages of 10 and 16. This paper identifies the social class-based practices that drive the influence of parental education on the development of political interest among early adolescents and explains why the social gap grows at this point. The paper draws on two panel surveys, the Citizenship Education Longitudinal Study and the Understanding Society Youth Survey, and applies latent growth curve modelling and path analysis. The findings show that university-educated parents influence the change in political interest of their children in early adolescence by raising their educational as-pirations, enabling their access to political activities in school, choosing the school for their children, taking their children to museums and art galleries and influencing their children's friendship groups.


INTRODUCTION
Parental educational attainment has repeatedly been found to be strongly associated with their children's political attitudes, including political interest (Flanagan & Levine, 2010;Kim & Lim, 2019;McFarland & Thomas, 2006;Verba et al., 2003).Moreover, it appears to be more influential than other indicators of social background, such as parental occupational status or household income (Janmaat & Hoskins, 2022;Lahtinen et al., 2019).Political interest is of particular importance as it predicts future political engagement (Shehata & Amnå, 2019;Van Deth et al., 2007;Verba et al., 1995).
There are significant social inequalities in political engagement in the United Kingdom, and these inequalities have been found to be transferred and socially reproduced from one generation to the next (Hoskins & Janmaat, 2019;Janmaat & Hoskins, 2022).Research has worryingly suggested that it is the differences in parental educational levels that are playing a role in the intergenerational transmission of socioeconomic inequalities in political engagement (Gidengil et al., 2016;Verba et al., 2003).According to this argument, the educational attainment of one generation is creating a pathway to socially reproduce existing inequalities in political voice in the next (Hoskins & Janmaat, 2019).This research suggests that more educated parents are engaging in certain socialisation behaviours and practices that lead to higher levels of political interest in their children (Hoskins & Janmaat, 2019).Nevertheless, identifying the social class behaviours and practices through which the transmission takes place between parents and children during the critical age period in which political interest develops has yet to be fully explored.
Analysis of nationally representative longitudinal data in England has shown that the key stage when political attitudes are malleable and when parental education begins to influence children's political interest is early adolescence (ages 10-16) (Janmaat & Hoskins, 2022).Before the age of 10, parental education does not appear to have a relationship with children's political interest but between ages 10 and 16, this relationship grows rapidly (Janmaat & Hoskins, 2022).After age 16, attitudes such as political interest-as well as the social gap in political interest-have been found to stabilise (Janmaat & Hoskins, 2022;Neundorf et al., 2013;Prior, 2010).As Janmaat and Hoskins (2022) focused their research on describing the development of this gap, why it opens up between ages 10 and 16 and why

Key insights
What is the main issue that the paper addresses?
The paper addresses the process of the social reproduction of inequalities in political power in the United Kingdom by identifying how highly educated parents develop the political interest of their children and in turn create access for certain groups within the next generation to the world of politics.

What are the main insights that the paper provides?
The findings show that university-educated parents influence the increase in political interest of their children in adolescence by undertaking actions related to their education, such as raising their educational aspirations, choosing their school, enabling their access to political activities in school and taking their children on cultural trips.the influence of parental education grows between these ages has not been explained.Considering the age range when political interest is malleable, the most likely explanation for the growing influence of parental education is differences in the way parents bring up their children according to their social class (Hoskins & Janmaat, 2019).This paper will attempt to identify the social class-based mechanisms through which more educated parents foster their children's political interest between the ages of 10 and 16.Although the emphasis in this paper is on the effect of parents, we do not deny the influence of other agents on children's political interest, such as schools.Our argument is more subtle: we contend that educated parents socialise their children in such a way that they make greater use of the civic learning opportunities provided by these agents.

LITER ATUR E Political interest
Political interest can be defined as 'the degree to which politics arouses a citizen's curiosity' (Van Deth, 1990, p. 278) and is manifested in a citizen as someone 'who at least cares about some issue of public concern' (Dostie-Goule, 2009, p. 406).Political interest has been identified as an important motivational force for future political engagement (Shehata & Amnå, 2019) and found to predict future participation in a wide range of civic and political activities (Van Deth et al., 2007;Verba et al., 1995).In addition, political interest is associated with following news and current affairs (Prior, 2007) and higher levels of political knowledge (Delli Carpini & Keeter, 1996;Luskin, 1990).It can therefore be understood as a key quality to be promoted among young people and an attitude that can facilitate access to future power, privilege and political voice.Despite its importance, there is little in the way of systematic research undertaken to identify and test a range of parental class-based practices that may enable their children to become interested in politics during their most malleable years as early adolescents (Campbell & Horowitz, 2016).

Educated parents and the development of political interest as a child's cultural capital
In order to identify the mechanisms through which educated parents evoke an interest in politics in their children, it is necessary to draw on the sociological theories of Bourdieu and his concept of cultural capital.Cultural capital is argued by Bourdieu to exist in three forms: the embodied state, the objectified state and the institutionalised state i (Bourdieu, 1986;Bourdieu & Passeron, 1977).The embodied state of cultural capital can be understood as attitudes and values children have developed throughout their upbringing, for example, through attending certain types of schools and participating in certain types of activities inside and outside school, with outcomes that are not overtly understood to have monetary value (Bourdieu, 1986).These embodied attitudes and dispositions form part of the expected norms of the particular social class, which in turn can be exchanged for privileges (economic capital) later and/or for becoming an accepted member of privileged circles (social capital) and therefore for maintaining social status in a family from one generation to the next (Archer et al., 2015;Bourdieu, 1986;Bourdieu & Passeron, 1977;Kisida et al., 2014).Examples of the development of cultural capital are family trips to art galleries, which have been found to embody in a child an appreciation of the arts, which is a disposition frequently associated with the middle and upper classes (Kisida et al., 2014).
Most research using Bourdieu's theory examines the effects of social class and cultural capital on the development of academic achievement and thereby on access to good careers and wealth (Archer et al., 2015).Nevertheless, as with recent research in this field (Kim & Lim, 2019;Kisida et al., 2014;McFarland & Thomas, 2006), we consider in this paper that political attitudes are a form of cultural capital developed within educated families through specific social class-based mechanisms that offer the next generation the potential for acquiring social status, privileges and potential access to political power and voice.In this paper we specifically consider political interest to be an embodied attitude that is developed by educated parents through specific class-based practices during their children's upbringing, for example, through school choice and family activities.We argue that political interest is not only an educated class-based norm but also has the potential to be exchanged for access to power and wealth.It is a form of cultural capital applicable in countries like the United Kingdom, where engaging in the formal political system is useful and relevant towards maintaining individual and class-based wealth and status, which we recognise may not be the case in all countries (see Kasara & Suryanarayan, 2015).Establishing the precise class-based practices in a UK context that are associated with the development of political interest is the focus of this paper.Theories of the development of cultural capital (Ball et al., 1996;Bourdieu, 1986;Bourdieu & Passeron, 1977) provide us with a basis to identify a set of mechanisms and develop hypotheses about the types of behaviours and practices of parents that could lead to increased political interest of their children.

MECHANISMS OF PARENTAL INFLUENCE
The home environment is typically where the process of embodying cultural capital is most salient.In a detailed class-based analysis of the socialisation process, Calarco (2018) established how more educated parents typically form 'negotiation households' in which children are encouraged to develop and express their own opinions, motivate these opinions with arguments and take part in discussions about broader social issues.These experiences in the home environment are therefore expected to be crucial in the development of both academic and civic skills, attitudes and dispositions (Schulz et al., 2010).
Experiencing positive social learning environments for the development of civic skills in the home is said to enable young people to seize civic learning opportunities at school (Hoskins & Janmaat, 2019).In turn, these opportunities-such as an open climate of discussing political and social issues in class and activities aimed at raising student voice-have been found to greatly foster political engagement (Barrett, 2012;Campbell, 2008;Torney-Purta, 2002;Verba et al., 1995).Among these activities it is the open classroom climate, in which students report the opportunity to express their opinions and have their voice listened to in discussions, that over the last 20 years has been the most consistently found method to increase students' political engagement (Campbell, 2008;Claes et al., 2017;Knowles & McCafferty-Wright, 2015;Kuang et al., 2018;Torney-Purta et al., 2001).The process of enabling students to understand the rules of the game on acceptable behaviour and know how to participate in a classroom debate is an example of embodied capital, which is learnt in the home and then used by students as a resource in the school environment.It works in much the same way as Archer et al. (2015) describes the case for skills that are learnt in the home and that go on to influence educational achievement.
In addition to seeking out civic learning experiences at school, young people with more educated parents are also more likely to seek civic and political activities outside of school.Theories based on cultural capital would suggest that at least in England, it is likely to be more educated parents who encourage their children to participate in civic activities (McFarland & Thomas, 2006).Based on the prior literature, we would suggest that participation in clubs and associations with a civic theme-such as care for the environment, human rights, political debate or community help-is likely to enhance the learning of political interest, and that this is sustained into adulthood (McFarland & Thomas, 2006).
Another learning outcome largely accounted for by middle-class socialisation in the home is argued to be the embodiment of general attitudes, such as the importance of education as a whole and the need to go to university (Kim & Lim, 2019).One of the attitudes that has consistently been found to be associated with political interest at age 13 is young people's expectations regarding higher education (Kim & Lim, 2019;McFarland & Thomas, 2006;Niemi & Junn, 1998;Torney-Purta et al., 2001).Lauglo (2011) has argued that the connection between educational aspirations and political interest is via family socialisation and the embodiment of the perceived need to participate in the public sphere.
Parents can also get involved in their children's development by taking them on cultural trips (e.g., to museums and art galleries) in the community.There is significant evidence that these types of experiences benefit education outcomes (Dumais, 2005) and there is some initial evidence of the effect of cultural trips on learning the skills that are relevant for political engagement (critical thinking, empathy and tolerance) (Greene et al., 2014).These cultural activities undertaken by parents have been identified as examples of the development of embodied cultural capital by multiple researchers (Byun et al., 2012;De Graaf et al., 2000;DiMaggio, 1982;Dumais, 2002;Jaeger, 2009;Kaufman & Gabler, 2004;Nagel, 2010).They can be understood as strategies employed by parents that help their children embody attitudes and dispositions associated with a high social class (Bourdieu, 1986).More recent research has begun from the premise that middle-class parents already undertake these activities when their children are very young and attempts to assess whether interventions such as school trips can compensate for the missed socialisation of children from disadvantaged backgrounds (Greene et al., 2014;Kisida et al., 2014).In the United States a largescale randomised control trial demonstrated that they could (Greene et al., 2014;Kisida et al., 2014).The fact that interventions can make a difference, and that these experiences are more effective for less advantaged students, is a nuance to the original more fixed concept of social class and cultural capital.
In addition to giving their children the political skills needed to participate in civic activities in school, there are a variety of other ways in which parents can influence the school experience of their children.In England the most striking of these is through choosing the school their children attend.School choice is argued to be influenced by the education levels and social class of parents (Ball et al., 1996;Burgess et al., 2011).Factors such as school quality, socioeconomic and ethnic composition of the students at the school and disciplinary climate are all factors that parents cite as contributing to their choice of school for their children (Burgess et al., 2011).Middle class, and thus more educated, parents typically go to great lengths, including moving home, to be in the catchment area for the desired school for their children.Consequently, they are more likely to accomplish their choice of school (Burgess et al., 2011).
It is unlikely that parents of any education level will select a school based on whether it is likely to increase the levels of political interest of their child.Nevertheless, the factors known to influence school choice are also likely to influence levels of political interest of the child.For example, as described above, the most educated parents are most likely to succeed at getting their children into schools that are ranked as the highest quality.Thus, these schools are dominated by more advantaged students.We know that the socioeconomic composition of the school influences the development of a range of civic competences over and above young people's own social background (Hoskins & Janmaat, 2019;Janmaat, 2022;Janmaat & Hoskins, 2022).School social composition is said to have this effect because it gives rise to particular peer cultures, conducive to or undermining political engagement (Ichilov, 2002;Jacobsen et al., 2012;Van de Werfhorst, 2007).
Influencing the choices of young people's friends is another way in which parents can influence political interest.The choice of where the family lives, the choice of the school and the choice of community involvement creates a pool of persons from which young people develop their friendship networks (Landini et al., 2016).Scholars have found that the potential peer groups available in certain schools can be prioritised even over markers of quality when parents are deciding on schools (Holme, 2002).Positive reinforcement of certain peers over others may also influence friendship groups (Keijser et al., 2012).In turn, friends' attitudes to politics have been found to influence political interest (Dostie-Goule, 2009) and societal interest more generally (Wanders et al., 2020).It may well be that parents are thus having an indirect effect on their children's political interest by influencing their peer group and friends.

Hypotheses
We have thus identified a series of seven potential mechanisms, from discussions in the home to friendship groups, through which educated parents positively influence the political interest of their children between the ages of 10 and 16.We group the hypotheses derived from these mechanisms according to the location of the actual activity (i.e., in the home, at school or in the community).Accordingly, we hypothesise that more educated parents engage in the following activities-or enable their children to pursue the following learning opportunities-that in turn lead to a higher level of political interest in their children.
In the home, such parents H1: have more academic and political discussions with their children; H2: encourage their children's desire and belief that they will go to university.At school, they H3: encourage their children to seize opportunities for practising and learning political engagement, such as taking part in open political discussions in class and participating in student voice activities; H4: choose schools for their children where more middle-class students attend.
In the community, they H5: encourage their children to use the opportunities for practising and learning political engagement; H6: take their children to cultural activities such as art galleries and museums; H7: ensure that their children are developing more politically engaged friendship groups.

DATA SOURCES
We explore how educated parents shape their children's political interest with two longitudinal data sources, the Citizenship Education Longitudinal Study (CELS) and the Youth Survey of Understanding Society (YSUS).Earlier we proposed that parents influence the political interest of their children through activities that occur in the home, school and community.While CELS contains useful measures of school activities and community activities, YSUS includes rich data on activities in the home and parent-led activities in the community that are likely to be conducive for young people's political interest.Thus, in their combination, the data sources allow us to explore the mechanisms of the influence of parental education in a more complete fashion.CELS was originally commissioned by the Department for Children, Schools and Families (DCSF) to assess the effects of citizenship education, which was introduced as a compulsory  (2002-2003, 2004-2005 and 2006-2007) were collected from a nationally representative sample of 75 state-maintained schools in England (Benton et al., 2008).Within the sampled schools, all the students of a school year were selected (i.e., all the students in a certain grade; a grade can include one or several classes).We used the data of these first three waves, covering an age range of 11 to 16 and comprising 6113 respondents who took part in all three waves.
Understanding Society and its predecessor, the British Household Panel Study, is the most prominent and longest-running longitudinal multi-panel study in the United Kingdom (Understanding Society, 2021).Since 2009, Understanding Society has tracked around 40,000 households (more than 100,000 individuals) by surveying them every year.Household members aged 16 and older are included in the Main Survey; those aged 10 to 15 are included in the Youth Survey.New members are included in these surveys when they turn 16 and 10, respectively.As the Youth Survey includes approximately 700 children at each age, we selected and pooled the data of 10 and 11-year-olds in 2011 (Wave 3) and 10 and 11-year-olds in 2013 (Wave 5) to achieve a sufficiently large sample size.We tracked this group every two years until ages 14 and 15 (Wave 7 for the oldest batch; Wave 9-the latest wave available-for the youngest batch).The group consists of 2355 respondents participating in each of three rounds (ages 10 and 11 = Round 1; ages 12 and 13 = Round 2; ages 14 and 15 = Round 3).
As with every longitudinal data source, both CELS and YSUS experienced attrition.This was most pronounced in CELS, with only 32.9% of the 18,583 respondents in Wave 1 taking part in all three waves.In YSUS the dropout rate was much less dramatic: of the 2521 respondents (aged 10 and 11) in the first round, 2355 (or 93.4%) were still in the survey in the third round.In CELS the attrition was somewhat selective on parental education, with those participating in all three waves reporting slightly higher levels of mother's and father's education than all respondents taking part in Wave 1 (e.g., while among the former 29.2% said their mother studied at university, 27.6% said so among the latter).No selective dropout was found based on parental education in YSUS.Importantly, in neither of the two data sources was there any bias in the attrition based on political interest, as the main dependent variable.Selective attrition is problematic when exploring aggregate levels of the variables of interest.However, when assessing relations between variables (which we are mainly interested in), selective attrition is unlikely to result in a bias when the analytical models include the variables on which the attrition is found (cf.Paterson, 2013).Since we found selective attrition on parental education, we will therefore include this variable in the analysis to meet this requirement.

VARI ABLES
Political interest, as our outcome of interest, is captured in YSUS with the item 'How interested are you in politics?' [response categories: 1 = not interested; 2 = fairly interested; 3 = very interested], which was included in every other wave starting from Wave 3 (i.e., Waves 3, 5, 7 and 9).In CELS, the equivalent item is part of a battery of items broadly tapping political efficacy.'How much do you agree or disagree with each of the following statements about politics?':'I am very interested in politics' [response categories: 1 = strongly disagree; 2 = disagree; 3 = neither agree nor disagree; 4 = agree; 5 = strongly agree].This item was asked in every CELS wave.In CELS we relied on respondents' reports about their mother's and father's highest level of education [response categories: 1 = left full-time education at 16; 2 = left after college; ii 3 = studied at university] to tap parental education, as our main predictor.Rather than tapping this predictor with a continuous variable, we created five dummy variables that identify the combined education experience of the parents, as this helps us to assess the impact of parental education in a more granular way.These dummy variables include (1) both parents left full-time education at 16; (2) one completed college, the other left at 16; (3) both parents completed college; (4) one completed university, the other left earlier; (5) both completed university.iii In the case of YSUS we use data on parental education as reported by the parents themselves in the Main Survey, which ensures greater accuracy.We merged this data with the Youth Survey and created a parental education variable in the same way as in the CELS data.The difference with the latter is that the dummy variables concern completed qualifications: (1) both completed lower secondary; (2) one completed college, the other completed lower secondary; (3) both completed college; (4) one completed university, the other has a lower qualification; (5) both completed university.
We further note that the two parental education variables have missing values.This is particularly the case for YSUS, as including parental education in the analyses based on YSUS data reduces the number of observations to 1164 (see Table 2) and 1213 (see Table 4).For CELS this fallout is less dramatic but the number of observations still goes down to 3888 (see Table 1) and 3894 (see Table 3).To check whether this attrition produces a bias, we carried out additional analyses with a series of dummies for parental education that also include categories combining valid information on one parent and missing information on the other (i.e., eight dummies in total).These robustness analyses, which are based on samples of 2223 for YSUS and 4593 for CELS and are presented in the Supplementary Data (S1), produce the same results as those reported below.
We argue that discussions in the home and the development of educational aspirations are activities that in the main take place in the home.iv The second of these is measured with the YSUS item 'Would you like to go on to do further full-time education at a college or university after you finish school?' [0 = no; 1 = yes], henceforth educational aspirations.Discussions in the home is captured with the YSUS item 'We discuss books at home' [1 = never << >> 4 = always].Although discussions about books could involve topics other than social and political issues, we felt that this item captured the aforementioned propensity of middle-class parents to encourage their children's participation in discussions.We further included controls for gender [0 = male; 1 = female] and ethnic identity [0 = other; 1 = White British].We did not add other controls because the two datasets do not have any other similar variables that could potentially be used as controls.Adding dissimilar controls would make the analyses less equivalent.
We created three variables in CELS to tap the relevant school mediating factors: school social composition (school SES), open climate of classroom discussions (open climate) and participation in school-based political activities (in-school political activities).School SES represents the school average of the respondents' scores on parental education from Wave 3 (i.e., from Year 11 when respondents are aged 15-16).It is the only school-level variable among the independent variables.Open climate was measured with a scale consisting of six items from Wave 3 tapping students' perceptions of how free the climate of discussions in a class is and teachers' facilitation of this (see Appendix A for the wording of these items).The scale has a robust level of internal coherence (Cronbach's alpha = 0.811) and represents the saved output of a factor analysis (i.e., the factor scores).Another composite variable was created to gauge in-school political activities.This variable represents the sum of the responses to four items from Wave 3 asking students whether they have participated in debates, a student council, elections for the school or mock elections in the last year.It has a minimum of 0 (not participated in any clubs or events) and a maximum of 4 (participated in all clubs and events).
We identified participation in out-of-school civic activities (civic activities outside school), parents taking their children to cultural activities and influencing friendship groups as community activities.Civic activities outside school is the sum of six CELS Wave 3 items asking students whether, in the last year, they have taken part in environmental groups, human rights organisations, debating clubs and religious groups, and whether they have helped in the local community or have raised money for a good cause or charity

ANALY TIC STR ATEGY
We start with trend graphs showing in a detailed way how political interest develops during early adolescence for children from different social backgrounds.Next, we explore the effect of parental education on the age 10-12 level of political interest and the age 10-16 change in political interest using latent growth curve modelling (LGCM) to assess whether the trends identified in the graphs are significant.It also helps us to establish at what point during childhood parental education leaves its mark and through which mechanisms it does so: a link between parental education and the initial level suggests an influence already in early childhood; a link between parental education and the subsequent change (and not with the initial level) suggests that parental education only starts to influence political interest in adolescence.We would expect to find evidence for the latter as Janmaat and Hoskins (2022) already found parental education to only be related to the change in political interest.
LGCM permits the modelling of the initial level and subsequent change in the dependent variable for individual respondents by estimating a random intercept and random slope, which are generated as latent variables.Subsequently, it can link independent variables to the intercept and the slope, which enables an assessment of the degree to which the interindividual variance in the initial level and in the subsequent change in the dependent variable can be accounted for by the independent variables (Bollen & Curran, 2006;Neundorf et al., 2013).In our case, a positive effect of parental education on the slope (i.e., change) of political interest would mean that the political interest of children from well-educated families has risen faster (or declined less steeply) than that of children with less educated parents.We will estimate the random slope (i.e., subsequent change) in political interest based on three time points in both the CELS and YSUS data (age 11-12, age 13-14 and age 15-16 in case of CELS; age 10-11, age 12-13 and age 14-15 in case of YSUS).v Three repeated measurements is the minimum number required for growth models such as LGCM (Curran et al., 2010).Owing to the nested structure of the sample in the case of CELS (respondents in grades; grades in schools), we carried out the LGCM analyses on this data with robust standard errors to correct for the bias produced by the non-independence of observations (cf.Muthen & Muthen, 2009).This is necessary to accurately estimate the effects of schoollevel variables, such as school social composition.
An additional benefit of LGCM is that it retains cases with missing values in one or more time measurements of the dependent variable.This is important as attrition affects all panel data, and CELS in particular, as noted before.Mplus, the software we use to run the LGCM analyses, preserves these cases by offering a maximum likelihood single imputation estimation (Neundorf et al., 2013).
Finally, we run a combination of LGCM and path analysis to assess to what extent the influence of parental education on the level and change of political interest runs through the mechanisms representing the home, school and community factors.Path analysis (sometimes called the structural component of structural equation modelling) allows for the identification of direct and indirect effects of an exogenous independent variable on the outcome, with 'indirect' referring to effects as mediated through other endogenous independent variables (Pearl, 2018).To assess these indirect effects, the analysis first relates the exogenous variable to the endogenous variables (or mediators) and then relates the endogenous variables to the outcome.It is these properties that make path analysis ideally suited for exploring the mechanisms of the influence of parental education as the main objective of the current paper.We consider these mechanisms as endogenous variables (mediators) between parental education and political interest.In the interest of transparency, we provide the Mplus syntax of the combined LGCM/path analysis for the model on the CELS data in the Supplementary Data (S2).
We further note that combining path analysis with LGCM puts us in a better position to address confounding influences when assessing the links between the mechanisms and political interest.The LGCM component allows us to automatically control for factors that drive both the initial level of political interest and the educational conditions proposed to influence political interest, including the aforementioned school mechanisms.In other words, this analysis addresses selection effects by incorporating them into the initial level of political interest (see also Finkel, 1995;Persson, 2012).Relating the school mechanisms to the change in political interest thus allows us to arrive at a more accurate assessment of their impact (i.e., net of selection effects).We cannot make absolute causal claims, however, as there is always a possibility that there are unmeasured factors that influence both the post-11 development of political interest and the school mechanisms we are assessing.

R ESULTS Trend analysis: A growing social gap in political interest
The results of the trend analysis of political interest for both the CELS and YSUS data show that although levels of political interest start off in a similar place for children of different social backgrounds, the major rise in political interest is for children where both parents have been to university (Figures 1 and 2).The political interest of all other children increases only slightly during early adolescence (a hardly perceptible increase with a dip in the middle for the trends based on the CELS data and a modest increase for those based on the YSUS data).In the CELS data, only the children with one parent who attended university have a marginally higher growth than those from less well-educated families.These minor differences aside, it is striking that the two datasets produce such similar trends, despite different measurements of the parental education variables and different modes and times of data collection.These results also echo those of Janmaat and Hoskins (2022), who similarly found the social gap in political interest to open up between ages 11 and 16.Our study improves on theirs by measuring parental education in a more granular way and by showing that it is the children with university-educated parents who move apart from the rest.As Janmaat and Hoskins (2022) used a data source from the 1990s (the British Household Panel Study) and we have used sources from the 2000s (CELS) and the 2010s (YSUS), we can be fairly confident in concluding that the growing social gap in political interest is an enduring phenomenon of British society.

LGCA and SEM models: Explaining the growing social gap in political interest
Tables 1 and 2, which present the results of an LGC model including only parental education and two control variables, confirm the patterns from the trend analyses: while parental education appears unrelated to the initial level of political interest, children from families where both parents went to university show a significantly steeper rise in political interest than those from families where both parents left education at age 16 (the reference category).In the CELS data, children where one parent went to university also show significantly higher growth.Thus, as was evident from the trend analyses, the social gap in political interest widens mainly because children with university-educated parents move apart from the other parental education groups.The high t ratios (β/SE statistics in the tables) for the 'both parents studied at university' effect on the change in political interest (5.5 in the CELS data and 3.5 in the YSUS data), moreover, show that the significantly higher growth for this group is quite strong.
We now turn to the combined LGC and path analyses to assess through which mechanisms parental education exerts its influence (see Tables 3 and 4).vi The first part of these tables shows the associations between the parental education categories and the variables that we proposed as mechanisms/mediators. Obviously, if these associations are not significant, parental education is unlikely to influence political interest through these mechanisms.The second part shows the direct effects of the parental education and mediator variables on the initial level (column 2) and subsequent change of political interest (column 3).It also shows the indirect effects of the parental education variables on the change in political interest (column 4).These indirect effects are presented as the sum of the effects through the mediating variables.We only present these indirect effects for the change in political interest because parental education is not related to the initial level of political interest.We emphasise that each table shows the results of one model.
Part I in both Tables 3 and 4 shows that the 'both parents went to university' category is significantly related to all the mediating variables for home, school and community.It shows a strong negative association with 'friends not interested in politics' and a fairly to very strong positive association with the other mediators.Thus, by comparison to the most disadvantaged group (i.e., the reference category), children with university-educated parents go to more prestigious schools, take more part in classroom discussions and civic/political activities in and outside school, have less politically disinterested friends, have stronger educational ambitions, more often discuss books at home and are more often taken to museums or art galleries by their parents.The 'one parent went to university' group shows significant associations with most mediators but not with open classroom climate and educational aspirations.The other parental education categories have less significant links with the mediators, particularly with those concerning home influences.Broadly, however, these patterns show that the education levels of the parents matter for all the circumstances that impact on political interest: children from well-educated families, particularly those with university-educated parents, have greater access to civic learning opportunities in school, encounter a more stimulating home environment and experience more opportunities conducive for engagement in the community.
Part II in both Tables 3 and 4 shows that parental education mostly influences the change in political interest in an indirect way (i.e., through the mediators).This can be concluded by comparing the direct effects of the parental education categories including their t values in column 3 to their indirect effects in column 4. For instance, in Table 3 the direct effect of 'both parents studied at university' is 0.10* (t = 2.6), while the sum of its indirect effects is 0.11*** (t = 4.7).Thus, the models of Tables 3 and 4 are relatively successful in identifying the pathways through which parental education influences the development of political interest during early adolescence.We can conclude that the widening social gap in political interest can largely be explained by the mechanisms that we have proposed: it is mostly through experiencing these conditions that children from more educated families move apart from those with more disadvantaged backgrounds in their political interest.This is not to say that we have captured all the mechanisms of the influence of parental education.Both tables show that there is still a direct-or we might say 'residual'-effect  3 also shows a direct effect of 'one parent studied at university'.These effects have been much reduced, however, by comparison to those presented in Tables 1 and 2: the effect of 'both parents studied at university' in the CELS data has gone down from 0.21*** (t = 5.5) to 0.10* (t = 2.6); the equivalent effect in the YSYS data declined from 0.15** (t = 3.5) to 0.11** (t = 2.6).Moreover, it is understandable that there are still these residual effects as neither CELS nor YSUS captured all the proposed mechanisms, given their emphasis, respectively, on the school and home environments.If there had been a data source allowing us to test all the mechanisms at once, we might have been able to 'explain away' the effect of parental education entirely.
It is further worth noting that the mediating variables also have a substantial independent effect on the change in political interest (i.e., an effect not instigated by parental education).This can be gauged by comparing the explained variances across the analyses with and without the mediators.Once the mediators are included, the explained variance jumps from 9.5% (Table 1) to an impressive 28.0% (Table 3) in the analyses on the CELS data and from 9.3% (Table 2) to 15.5% (Table 4) in the analyses on the YSUS data.These percentages of explained variance are rare in analyses of attitudes and testify to the relative success of the mediators in explaining the change in political interest.Among these mediators, school SES, in-school political activities, friends not interested in politics, trips to museums and art galleries and educational aspirations are quite strong predictors, with effects at the 0.01 level of significance or less.vii Only civic activities outside school and discussing books at home are not related to the change in political interest.These mediators, however, are associated with the initial level of political interest, which suggests that they have an influence in early childhood, rather than adolescence.

DISCUSSION
In this paper we are mostly interested in the change in political interest between ages 10 and 16 and the mechanisms through which educated parents influence the political interest of their children during this life stage.This is because our results suggest no effect of parental education on initial levels of political interest at age 10.In contrast, we found parental education to have a marked effect on the development of political interest during early adolescence: children with university-educated parents showed a steeper growth in political interest than other groups, resulting in larger social differences in political interest at age 16.Thus, similar to Janmaat and Hoskins (2022), but using different data sources and more precise measures of parental education, we found the social gap in political interest to widen between ages 10 and 16.
The original contribution of this paper lies in identifying the parental mechanisms that create the social gap in levels of political interest in children between the ages of 10 and 16.Overall, regarding the change in political interest, we find that five out of the seven mechanisms proposed by our hypotheses capture practices that parents with higher levels of education (particularly those with degrees) engage in, and that in turn enhance their offspring's levels of political interest during this critical development period.In fact, our path analyses showed that these mechanisms could account for most of the effect of parental education on the change in political interest.In other words, they can largely explain why the social gap widens during early adolescence.
In the home, these parental practices concerned encouraging children to aspire towards university education.Bridging home and school, it is the school that parents send their children to that is crucial.Sending a child to a school that predominantly contains other children from highly educated parents plays an important role in developing their child's political interest.It needs to be noted here that our school data (CELS) is from more than 15 years ago and that school choice has expanded since then as a result of the Academy Act of 2010 and the entry of free schools into the education market the following year (Bhattacharya, 2013).This could have resulted in greater social segregation and a stronger effect of school social composition on children's levels of political interest.In addition, educated parents are enabling their children to access more of the political activities inside these schools.Parents also appeared to be having an indirect impact on their children's political interest by influencing their children's friendship groups more broadly.Finally, parents taking their children to art galleries and museums was another way in which educated parents cultivated an interest in politics in their children.Although discussions of books at home appeared not to be related to the change in political interest, we did find it to have an effect on initial levels of political interest, which suggests that these discussions may well provide the initial learning environment for political interest (Schulz et al., 2010).
Encouraging high educational aspirations and having discussions in the home is in line with theories of the development of cultural capital.The fact that we found parental education to also be strongly connected to the civic learning opportunities in schools suggests that children from well-educated families use the skills and attitudes learnt in the home environment to access these opportunities, which would be in agreement with Bourdieu's (1986) theory of the embodiment of cultural capital and how this cultural capital can then be used to their advantage.
Taking children to museums and art galleries is an example of cultural capital in action.Young people learn about and discuss the world through engaging with their family in cultural activities and, as Bourdieu suggested, attitudes such as political interest are embodied through undertaking these activities.Taking children to museums and art galleries is clearly attributed to privileged social classes (Kisida et al., 2014).The effectiveness of this strategy in increasing levels of political interest provides considerable support for Bourdieu's theory of the process of embodiment of cultural capital.Observing evidence of Bourdieu's cultural capital theory in action in increasing the political interest of middle-class children, and thereby offering them a route to exercising voice and power, provides an extension to his initial writing that focused on the embodiment of cultural capital in terms of the appreciation of art and aesthetics as a demonstration of social class.It could also be that the declaration of political interest of children in the survey responses is, in and of itself, a part of the embodiment of becoming middle class.Working-class students may have interests in areas that could be understood as political, but these students would not classify these concerns in terms of having an interest in politics.
There is little within the existing cultural capital literature on the influence of parents on their children's potential peer and friendship groups.As well as finding that parents had a general influence on their child's peers and that this affected their children's levels of political interest, school social composition (as reflecting parental school choice) had an additional effect on this process.These results build on the broader empirical literature on the effect of the loss of truly comprehensive education through the implementation of school choice policies (Ball et al., 1996;Burgess et al., 2011).
It is necessary to acknowledge again at this point that our results demonstrate the process of embodiment of political interest as a form of cultural capital in the context of a specific UK democratic political system with a history of neoliberal governments maintaining relatively high levels of economic inequality.We could speculate that what counts as political cultural capital may differ in different country contexts and that different results may be found in more egalitarian democratic countries, in authoritarian countries and in some developing countries, where the administrative systems are less able to bring in taxation and control wealth (Kasara & Suryanarayan, 2015).It might also be the case that in countries with substantial trade union membership, trade unions could act as an equalising force in training the next generation from less educated families to take up an interest in politics.
It is more difficult from the literature to explain why two out of the seven mechanisms were not effective at increasing political interest between ages 10 and 16.This may be because both these variables were found to influence the initial levels of political interest.The first of these, discussing books at home (YSUS data) was found to influence the initial levels of political interest.It was the only measure available on discussions in the home in the two datasets.Perhaps this result shows that discussing books has an earlier effect on political interest and one that dissipates after the age of 10.Alternatively, it could be due to the measure strictly capturing the content of the books discussed and not political discussions in the home or being involved in decision-making in the home, which might have made the item more relevant to changes in political interest between the ages of 10 and 16.
The second mechanism that does not appear in the full model to be associated with young people's changing levels of political interest between age 10 and 16 is participating in political activities outside of school.Interestingly, this variable is also associated with the initial levels of political interest suggesting, rather surprisingly, that young people are already engaged in these activities at the age of 11 (the initial age for the CELs data) and had already developed their political interest from this.There are still outstanding questions to be addressed in understanding the socialisation process that can be answered when better quality measures are available.Notably, we need more comprehensive data allowing us to measure and assess how children gain and capitalise on the cultural capital learnt at home to access and reap the benefits of civic learning opportunities at school and in the wider community.

CONCLUSION
This paper has combined the strengths of two panel studies that were run in England to identify the mechanisms through which parental education levels translate into politically interested children during the critical period of their early adolescence.The theory of the embodiment of cultural capital has helped us to identify social class-based behaviours and actions undertaken by highly educated parents, and we found five of these able to account for the effect of parental education on the changing levels of political interest of early adolescents.We have thus, at least in part, resolved some of the questions on how and why the gap in political interest between children of well and less well-educated families grows between ages 10 and 16.
The next step is to identify policies that can be put in place to rebalance the opportunities of children with less educated parents.The most obvious of these are school policies, as it is easier to intervene in schools than in the home.First, comprehensive education needs to be really comprehensive and bring students of diverse backgrounds together.This may mean removing school choice for the benefit of the common good.Second, schools need to organise more school trips to museums and art galleries, in particular schools recruiting mainly students from disadvantaged backgrounds.This has been found to be effective in the United States in terms of fostering attitudes similar to political interest (Greene et al., 2014).Third, teachers need to ensure that all social groups are involved in political activities in schools, so that all young people have the chance to learn that politics can be interesting for them.In this regard, teachers can make a difference in addressing inequalities in political interest if given the necessary training.Including students from all social backgrounds in political activities at school has been found to be effective in England (Hoskins & Janmaat, 2019).If all these steps were to be introduced in schools, there is a good chance that we could start to rebalance democracy so that all voices-regardless of social background-can be heard.

C O N F L I C T O F I N T E R E S T S TAT E M E N T
The authors declare no conflict of interest.

D ATA AVA I L A B I L I T Y S TAT E M E N T
Data are subject to third-party restrictions.

E T H I C S S TAT E M E N T
This paper uses public secondary data analysis only.

E N D N O T E S
i Objectified capital refers more to objects such as fine art.Information on these objects in the home is less available in our data sources.Institutionalised cultural capital refers to educational qualifications.In the English education system, most young people will not yet have received any qualifications by the age of 15, the age that our data analysis stops.ii In the English education system, college refers to sixth form college and will be the category selected if either of their parents undertook A levels or equivalent vocational qualifications (usually until the age of 18).iii We applied a gender-neutral approach regarding the measurement of parental education as we focus on the development of political interest among children from different social backgrounds.Had we been interested in exploring gender differences in political interest, then we would have distinguished between the effects of mother's and father's education as the literature has found these effects to differ for girls and boys (Gidengil et al., 2011;Jennings & Langton, 1969).Another reason why we applied a gender-neutral approach is that splitting parental education up into a mother's and father's component would have created too many categories, making the analysis of the social gap in political interest unmanageable.iv We note that educational aspirations are not only fostered through family socialisation.Some schools make considerable efforts to enhance the educational aspirations of students from disadvantaged backgrounds and, consequently, the perceived positive school atmosphere can make a difference to educational aspirations (Madarasova Geckova et al., 2010).v We performed these analyses assuming a linear growth trajectory.We investigated whether non-linear growth models would perform better in terms of the model fit statistics but they do not.Hence we opted for linear growth models.The non-linear growth models can be obtained from the authors upon request.vi We decided to present the results of these analyses in a table rather than in a diagram showing causal pathways because the multitude of relations would make the latter unreadable.vii These mediators are likely to be interrelated as it has, for instance, been found that schools with a more privileged intake offer more opportunities for student voice (McFarland & Starmanns, 2009).However, even if, say, school SES has taken away some explanatory power of in-school political activities and open climate, these two mediators still show an independent positive effect on the change in political interest (see Table 3, Part II, column 3).

14693518, 0 ,
Downloaded from https://bera-journals.onlinelibrary.wiley.com/doi/10.1002/berj.4002 by University College London UCL Library Services, Wiley Online Library on [16/04/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License curriculum component in 2001.It includes data from a cohort of youngsters who were aged 11 and 12 (Year 7; first year of secondary school) when they were surveyed for the first time (in 2002-2003).This group was re-surveyed every two years until 2010-2011 (aged 19 and 20) and again in 2014 (aged 23) (the last wave).The data for the first three waves 14693518, 0, Downloaded from https://bera-journals.onlinelibrary.wiley.com/doi/10.1002/berj.4002 by University College London UCL Library Services, Wiley Online Library on [16/04/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License [0 = not taken part in any of these clubs << >> 6 = taken part in all of them].Parents taking their children to cultural activities is measured using the YSUS item 'My parents take me to museums or art galleries' [1 = never; 2 = rarely; 3 = sometimes; 4 = often].Friendship groups is tapped with the CELS Wave 3 item 'My friends are not interested in politics' [1 = strongly disagree; 2 = disagree; 3 = neither agree nor disagree; 4 = agree; 5 = strongly agree].

14693518, 0 ,
Downloaded from https://bera-journals.onlinelibrary.wiley.com/doi/10.1002/berj.4002 by University College London UCL Library Services, Wiley Online Library on [16/04/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License The effect of parental education on the initial level and subsequent change of political interest: Results from an LGCM analysis based on YSUS data..1.*p < 0.05.**p < 0.01.14693518, 0, Downloaded from https://bera-journals.onlinelibrary.wiley.com/doi/10.1002/berj.4002 by University College London UCL Library Services, Wiley Online Library on [16/04/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License T A B L E 3 Direct and indirect effects of parental education on political interest: Results from a model combining LGCM and path analysis based on the CELS data.Part I: Effects of parental education on the mediating Both left education at age 16 (ref.)14693518, 0, Downloaded from https://bera-journals.onlinelibrary.wiley.com/doi/10.1002/berj.4002 by University College London UCL Library Services, Wiley Online Library on [16/04/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Part II: Direct and indirect effects on the initial level and subsequent change of political Downloaded from https://bera-journals.onlinelibrary.wiley.com/doi/10.1002/berj.4002 by University College London UCL Library Services, Wiley Online Library on [16/04/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License T A B L E 4 Direct and indirect effects of parental education on political interest: Results from a model combining LGCM and path analysis based on the YSUS data.Part I: Effects of parental education on the mediating Downloaded from https://bera-journals.onlinelibrary.wiley.com/doi/10.1002/berj.4002 by University College London UCL Library Services, Wiley Online Library on [16/04/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Part II: Direct and indirect effects on the initial level and subsequent change of political only a mediocre but still acceptable model fit.However, the equivalent LGC model without SEM has an RMSEA value of 0.02, indicating an excellent model fit (these results can be obtained from the authors upon request).T A B L E 4 (Continued) 14693518, 0, Downloaded from https://bera-journals.onlinelibrary.wiley.com/doi/10.1002/berj.4002 by University College London UCL Library Services, Wiley Online Library on [16/04/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License of 'both parents studied at university' on the change in political interest and Table

F
This work was supported by a Nuffield Foundation Grant for Research, Development and Analysis (No. 44157).

Initial level at ages 11 and 12 (intercept) Change between ages 11/12 and 15/16 (slope)
The development of political interest by parental education level (CELS data).The effect of parental education on the initial level and subsequent change of political interest: Results from an LGCM analysis based on CELS data.