Abstract
This study aims to contribute new insights into the way ‘political labour’ is divided in the household. I use data from a large-scale panel study, the Parent–Child Socialization Study 2012–2013, conducted among adolescents and both their parents in Belgium, to analyse the different ways in which family members engage in politics and influence each other’s political preferences. First, I analyse differences in political engagement between fathers, mothers and adolescents. Second, I present a full triadic structural equation model to measure the political influence that fathers, mothers and adolescent children exert on one another. The findings suggest that fathers are (still) more engaged in politics, but when it comes to preferences for political parties, both parents influence their partners and their adolescent children in equal measure.
Keywords
Introduction
Party attachments are not formed in isolation. Friends, colleagues and – most importantly – family members have been found to influence preferences for political parties. The influence of family members was the subject of numerous studies in the 1960s and 1970s, with the focus in electoral research later shifting to short-term voting motives such as campaigns, personalities and economics. In recent times, the debate around the social logic behind party attachments has received renewed attention, with new empirical work focusing on the applicability of the traditional social learning mechanisms to the current political culture. It is important to understand how these social learning mechanisms, rooted in a historical context of strong embeddedness in one single political environment (e.g. a Liberal or Conservative family), apply to the current, individualized political landscape. Earlier findings have been revised, while important advances have also been made in the study of genetic political parent–child inheritance. Thanks to the availability of large-scale panel datasets and more advanced analytical models, many empirical advances have also been made. These advances have allowed researchers to move beyond generalities and to dig deeper into the particular mechanisms behind the social network influence of political behaviours and attitudes.
This study aims to contribute to this debate by examining whether classical social learning mechanisms at the level of party preferences in families can still be applied in a de-aligned political context. I make grateful use of new empirical panel data, gathered in a fragmented multi-party setting (Belgium), that allows us to understand the different ways in which different family members influence each other’s party preferences over time. My stated goal is to further disentangle the differences between mothers and fathers in this political socialization process, a topic that is often disregarded due to data restrictions. Not only party alignments but also household patterns have changed dramatically since the first empirical studies on political socialization were published in the 1960s and 1970s, and I therefore aim to update the classical social learning mechanisms within this new societal and political context.
I particularly focus on the differences between fathers and mothers in this socialization process (Zuckerman et al., 2007). Most recent studies (Jennings et al., 2009) have paid little attention to the different roles fathers and mothers can play in the development of party preferences. This is remarkable as developmental studies indicate that mothers and fathers play different parenting roles and that they emphasize different political issues (Collins and Russell, 1991; Berelson et al., 1954). I aim to clearly disentangle the different political roles that mothers and fathers take on in the household and analyse the way in which they influence their children’s party preferences.
Empirically, I aim to further build on recent developments, among others by taking into account the fact that children – in spite of their age and limited experience – should not be treated merely as passive receivers of political stimuli (McDevitt and Chaffee, 2002; Zuckerman et al., 2007). In the new approach to family influences, family communication and socialization are seen as reciprocal processes, in which adolescents can initiate political discussions and incite parents to increase their civic competencies, and no longer as top-down processes of ‘inheritance’ (Glass et al., 1986), ‘imprinting’ (Niemi and Jennings, 1991) or even ‘family indoctrination’ (McCloksy and Dahlgren, 1959). I will take this reciprocal influence into account by applying structural equation modelling to a recently gathered large-scale panel dataset on parent–child interaction in Belgium, the Parent–Child Socialization Study (PCSS) 2012–2013 (Hooghe et al., 2013).
Studying social learning in a new political context
Theoretically, studies on parent–child transmission depart from a social learning perspective (Bandura, 1977). This ‘social logic of politics’ strand of research (Zuckerman, 2005) was crucial to early partisanship studies (Campbell et al., 1960). Subsequent empirical studies have indeed confirmed that party attachments generally develop early in life within one’s immediate social network (Jennings and Niemi, 1981; Jennings et al., 2009; Kroh and Selb, 2009).
Although this social learning model was very influential for the first generation of electoral studies, subsequent researchers substantially shifted away from this approach (Zuckerman et al., 2007). The focus on socially embedded partisanship to explain voting behaviour moved to more individualized explanations, such as candidate evaluations (Van der Brug, 2010), retrospective evaluations (Fiorina, 1978) and policy preferences, a logical development against the background of partisan de-alignment. Modes of participation have altered as well, with a stronger focus on non-institutionalized forms of political participation that particularly appeal to younger citizens such as protesting, online activism, lifestyle politics and consumer activism (Inglehart, 1997; Norris, 2002; Verba et al., 1995). All these evolutions have changed citizens’ formal and psychological attachments to political parties. One of the clearest consequences of these processes is the rise in electoral volatility (Dassonneville and Dejaeghere, 2014).
This is important, since clearly observable cues (e.g. ‘in this family, we are Liberals’) are essential in social learning theory (Bandura, 1977). For social learning to take place, those cues need to be regularly communicated to be retained, internalized and reproduced. But as party attachments have significantly changed over the past four decades, it is likely that these socializing cues, and therefore the complete process of social learning, have been affected as well. The question then becomes: to what extent are the social learning mechanisms that were applied to explain the early development of party attachments in a society with high partisanship levels (such as the US setting in which most studies were conducted) still applicable in a context where this self-perceived group membership has increasingly been losing appeal?
Evolution in the political roles of mothers and fathers
Studies published in the heyday of political socialization research in the 1960s and 1970s still guide our current expectations around the influence of household interaction on politics. However, subsequent decades were marked by major developments, for instance, at the level of the composition of household patterns. Therefore, the first central empirical goal of this paper is to update these previous findings to the current timeframe. In 1971, Jennings and Niemi had already analysed the differences in the political roles that mothers and fathers take on, and tested the supposed notion that fathers are the representatives of the political community and act as a bridge between the family and broader society (McClosky and Dahlgren, 1959). The results did not fully confirm their expectations. Although fathers have a persistent advantage when it comes to indicators of politicization, the researchers also observed an evolution towards a role-sharing division of political labour in the family. Jennings (1983) later confirmed that fathers have a stronger role as a source of political information and tend to act more as a referent, particularly for their sons.
In light of the recent research around participation, this gender gap within the family is not that surprising. At the level of conventional political participation (e.g. party membership, supporting candidates), gender is still one of the main sources of inequality (Burns et al., 2001), although studies have suggested that this gender gap has been reversed for voter turnout and other less institutionalized forms of participation (Marien et al., 2010). For the other variables in Jennings and Niemi’s approach to political labour (political efficacy and interest), empirical studies also point in the direction of a persistent gender gap (Bennett and Bennett, 1989; Burns et al., 2001; Marshall et al., 2007). In a first exploratory phase, I aim to develop an up-to-date picture of this division of political labour within the household.
The political influence of mothers and fathers
Although gender might be a determinant of political engagement within the family, this does not imply that this has immediate consequences for the political influence that parents exert on their children. This has been confirmed by large-scale studies conducted in the past decades. Jennings and Langton (1969) argued that a child is more likely to agree with the parent with whom he/she has a more ‘intense’ relationship, regardless of whether that parent is the mother or the father. A tendency in recent literature is to focus on a gender-specific instead of a gender-dominant transmission process (e.g. Nieuwbeerta and Wittebrood, 1995) in which sons are more heavily influenced by their father, and daughters by their mother. The idea is that children more strongly identify with a parent of the same gender so that father–son and mother–daughter transmission patterns are likely to be more clearly observable. Empirical evidence for this gender-specific hypothesis is of a mixed nature and strongly dependent on the political attitude being studied (Kulik, 2002; Roest et al., 2010). A recent study by Filler and Jennings (2015), for instance, showed that this gender-specific socialization pattern was the most dominant one for the transmission of gender role attitudes, and found an explanation for this in the fact that the attitudes studied in their work were explicitly gender-related. Although I focus on less gender-dependent variables in this paper, it is still useful to acknowledge this possibility and to take it into account.
In terms of party preferences, Zuckerman et al. (2007) made one of the most comprehensive recent studies to tackle the question of paternal versus maternal influence. They confirm the earlier findings of Jennings and Niemi (1974) and find mothers to be more at the centre of family politics, with the father occupying a more complex position. They explain the stronger influence of mothers by pointing out that children spend more time with their mothers, feel closer to them and therefore tend to be more influenced by them. I formulate my expectations vis-à-vis top-down parental influence based on these findings in Hypotheses 1 and 2 (H1 and H2) below.
When studying the influence between family members, one needs to take into account more than just the top-down processes. It also essential to look into the mutual influence between adult partners, and the trickle-up relation between adolescents and their parents. Mutual partner influence in particular has received more scholarly attention in recent times. On the one hand, I expected a strong correlation between the political preferences of partners to be the result of a mutual selection process (Alford et al., 2011). Individuals do not select a partner based on political party preferences, but the similarity is often a ‘by-product’ of the similarity between partners in other respects (Coffé and Need, 2010). Klofstad et al. (2012) have shown that spouses tend to share political predispositions, but that the latter do not play a role in the ‘selection of potential dates’. In ‘The Politics of Mate Choice’, Alford et al. (2011) also underscore the tendency of individuals to initially select (politically) like-minded mates, rather than assimilating their views to one another in the course of the marriage. On the other hand, empirical studies have shown that partners exert a strong mutual influence on each other because they live together and change opinions together (Zuckerman et al., 2005). Husbands and wives tend to have similar preferences prior to marrying (Watson et al., 2004), but this similarity grows stronger as time passes (Zuckerman et al., 2005). Coffé and Need (2010) indeed confirm that mutual influence is an important explanation for partner similarity.
Following these empirical findings, the political similarity between partners can be expected to be both the result of (1) general shared predispositions, and (2) a process of mutual socialization. Because of these two conflicting mechanisms, I will leave the process of mutual influence between marital partners open as an empirical research question, since no single hypothesis can be drawn from these inconsistent empirical findings.
Finally, political socialization scholars have underlined that the traditional top-down parent–child socialization approach needs to be revised and have instead proposed a more interactive family learning model in which children are not mere receivers of political stimuli. McDevitt and Chaffee (2002) here made a major contribution by showing that children can play a role in ‘parental growth’ – not so much by causing actual changes in their parents’ preferences, but rather through behaviour that increases their political engagement. They do not argue that children directly influence the political opinions of their parents and thus simply reverse the causality, but claim instead that they can initiate discussions and ‘act as a catalyst that shakes up the family system’ (295). Zuckerman et al. (2007) also stated that it is useful to take into account the reciprocal relationship between children and parents, since we cannot simply assume that socialization is a top-down process in a more contemporary approach to political socialization. As is evidenced by Hypothesis 3 (H3), I do not expect adolescent children to directly influence their parents, but I do take the possibility and the reciprocity of the relationship into account in light of these studies.
Data: PCSS 2012–2013
One of the common limitations of earlier studies was that the analyses were based on static data. In order to make reliable statements on the direction of influence between family members, however, we need a longitudinal approach. The panel data from the Belgian PCSS 2012–2013 were very useful in addressing the shortcomings of these earlier studies. This survey was administered among respondents in the Flemish part of the country, one of the most fragmented multi-party systems in Western Europe and one in which a large number of new and smaller parties have risen over the past decades (Deschouwer, 2009). Considering the high level of volatility in this party system, I expect this two-wave survey to allow mapping of the changes between the two waves.
The survey was conducted in 2012 and 2013 among 15- and 16-year-old adolescents and their parents in Belgium (Hooghe et al., 2013). In the first wave, a representative sample of 3,426 adolescents was interviewed during school hours via a written questionnaire. They simultaneously received a questionnaire for both their parents to complete at home.
The pupils were selected using a stratified random sample of 61 Dutch-language schools in Belgium, based both on province and school stream. Both parents submitted completed questionnaires for 60.8% of all adolescents; at least one of the two parents completed the survey for 72.7% of the adolescents. Both at the level of gender and school stream, the sample closely resembles the normal population distribution and can be considered representative for this age group in Flanders.
One year later, the researchers again visited the same schools. They were able to poll most of the adolescents who also participated in the initial survey. Of the original triads, 68.7% are also available in the second wave. From the original 3,426 adolescents, I have full information on 2,085 father–mother–child triads. From these 2,085 cases, I have full panel information on 1,430 (68.7 %) triads. In the first – descriptive – analysis, results are shown for Wave 2 variables for these 1,430 triads. I did not use listwise deletion in the full structural equation model because I wanted to make optimal use of all the available information, including that for the incomplete panel triads.
As my focus is on the influence between mothers, fathers and children within the household, I only selected those adolescent cases where all family members lived in the same household. This sample was used for all the descriptive analyses and for the structural equation model. 1 In terms of the representativeness of the sample, I compared the sample of adolescents with married parents in this study to the population and the full dataset. The slight underrepresentation of pupils with a lower socioeconomic status (SES) enrolled in vocational education becomes somewhat bigger when I select only those where both parents live in the same household.
Analyses
First, I aim to form a clearer view of the general political dynamics in contemporary households with a few basic indicators. 2 The first one is political interest, which was measured using a single indicator: ‘To what extent are you interested in politics and social issues?’, ranging from ‘ not interested’ (1) to ‘very interested’ (4). Following the news was measured through the question: ‘How often do you read, watch or listen to the news?’, ranging from ‘never’ (1) to ‘daily’ (5). The index participatory activities is a sum of three political activities: on a scale ranging from ‘never’ (1) and ‘from time to time’ (2) to ‘often’ (3), the adult respondents indicated how often they had ‘Looked up information on a party before going to the polling station’, ‘Helped a candidate during the election campaign’ and ‘Ran as a candidate for the elections’ during the previous electoral campaign. As mentioned above, these are three participatory activities that might be labelled conventional political participation. I selected these three to stay as close as possible to the original Jennings and Niemi (1971) index on conventional political participation (see below). In the PCSS data, there was no question on voting itself, since Belgium has a compulsory voting system and an average turnout of around 90%. Political efficacy was measured using a single indicator for internal efficacy: ‘How often do you think politics is so complicated that you do not understand what is going on’, ranging from 1, ‘most of the time’, to 5, ‘never’. The adolescent respondents were asked the same questions, but with respect to their intended behaviour as adults.
I compared the scores for these variables with a number of rudimentary calculations Jennings and Niemi (1971) made based on the 1965 University of Michigan data. These data are known as the Youth-Parent Socialization Panel Study, of which the first wave was conducted among a national sample of parents and children in the United States in 1965. The authors proposed a measure for the father’s net advantage over the mother for politicization indicators and subsequently calculated the cases in which fathers scored higher on an indicator than mothers, and vice versa. The cases in which mothers scored higher than fathers on a variable were then subtracted from the cases in which the fathers scored higher than the mothers to reach a ‘net advantage’ score, a procedure that I replicated with the PCSS data. Some indicators were measured in different ways in the two surveys and the results should consequently be interpreted with caution. But as I am analysing the differences between partners within one sample, and not comparing scores between the two samples, this should not result in major biases in the actual results.
Surprisingly, the results are comparable to those 1960 situation (see Table 1). Fathers are still more efficacious and interested, and they follow the news more often than their partners do. We do not see a clear difference for the participatory activities. While Jennings and Niemi (1971) found that fathers were 18% more likely to participate in electoral activities than mothers, we do not see a difference in the 2013 PCSS study, an observation in line with recent studies on turnout (Marien et al., 2010).
Fathers’ ‘net advantage’ over mothers for politicization indicators.
Note: N Follow news = 1,276; N Interest = 1,383; N Efficacy = 1,274; N Participatory activities = 1,224.
Reciprocal influence on voting intentions
Next, I examined the difference in political influence between fathers and mothers by mapping changes over time. I expected these short-term changes to be observable in this competitive multi-party setting, but to make sure these switches actually occurred, I mapped all movements between the two waves (the presentation of all party switches was added to online Appendix 3). Of the adolescents in the PCSS sample, 43.2% did not have a stable preference between the two waves; 25.6% of the fathers changed parties, while for the mothers this figure was 28.7%. In terms of switches in the direction of other family members, 20.3% of the mothers with a different party preference to their partners in Wave 1 changed their voting intention to that of the fathers in Wave 2, while 19.8% of the fathers in this situation changed their voting intention to that of the mothers in Wave 2. Of the adolescents whose voting intentions differed from those of both their parents in Wave 1, 10.4% switched to the party of their father, while 16.7% switched to the party of their mother.
It is clear that movements do indeed take place as family members adapt their voting intentions to those of other members of the family, but to fully grasp family members’ influence on each other, it is more useful to analyse an earlier step in the voting decision-making process: the ‘consideration stage’ (Bochsler and Sciarini, 2010). In the descriptive statistics, I only took into account actual changes in voting intention, but as the second wave was conducted only a year later, it is important to analyse subtler changes. I therefore do not analyse the actual changes in voting choice, but the subtler changes in the propensity to vote (cf. discussion on this measure below).
I estimate a structural equation model to analyse the reciprocal influence of adolescents, mothers and fathers on each other. This is a change model in which I include both cross-lagged effects and lagged dependent variables (see Figure 1). I recognize that including autoregressive terms can cause incorrect interpretation of other independent variables, particularly in mixed models (Achen, 2001; Plümper et al., 2005). The general idea behind the lagged dependent variable model is that the lagged value of y serves as a proxy for any unobserved between-person variance (Vaisey and Miles, 2014). An undesirable side effect of this procedure is that the autoregressive term can become very dominant in the model and decrease the effect sizes of other independent variables to such an extent that it can lead to an underestimation of the importance of these variables. However, the question of whether or not to include the lagged dependent variable in the model is not one of model fit, but rather a theoretical one.

Two-wave structural equation model predicting propensities to vote.
For this study, including the lagged dependent variable was theoretically important since I also aimed to compare the stability of the coefficients for mothers, fathers and adolescents in the sample. Comparison of these stability coefficients provides useful information on the likelihood of family influence. If the lagged y-coefficients are, for instance, stronger among adults than among adolescents, we can expect there to be less ‘room’ for external influence from other family members.
In the analyses, I focused on the four largest parties in the adolescent sample: the Christian Democrats (CD&V), the Flemish Nationalist Party (N-VA), the Green Party (Groen), and the radical right party, Vlaams Belang. I also applied these analyses to the other, smaller parties, and the results went in the same direction. But to guarantee the reliability of the presented results, I limited my selection to the four largest parties (n>300).
To test the gender-specific socialization mechanisms I described in the introduction, I ran additional multi-group models comparing male and female adolescents.
Dependent variable: propensity to vote
Although the bulk of electoral studies use voting intention or partisanship as a dependent variable, this methodological approach shows some important disadvantages, particularly in the case of multi-party systems. By forcing respondents to choose one party at a time, it becomes impossible to take into account the possibility of multiple party identifications (Weisberg 1999, 1980). By focusing only on actual voting intentions, it also becomes impossible to take into account possible changes to the actual decision due to strategic motives (Cox, 1997). In general, a large part of the decision-making process remains underexposed: we do for instance not know which other parties are under consideration. It is exactly this process that we can analyse with a measure for the propensity to vote for each of the different parties. All respondents were asked to indicate the likelihood they would ever vote for the stated parties in the future on a 0 to 10 scale. 3 In this way, it became possible to analyse how voters rated all parties, and not just the party they said they would vote for (Van der Eijk et al., 2006). This measure has been used in recent electoral research (Bochsler and Sciarini, 2010; Van der Brug, 2010) and has produced qualified results by splitting the party choice process into a consideration stage and the ultimate vote choice. It is precisely in this consideration stage, rather than in the actual voting decision, that I expect the first influence of family members to manifest itself. An additional empirical advantage is that this approach maximizes response rates for all analysed parties, which redresses the potential incomparability of results (van der Eijk et al., 2006).
Independent variables
As this analysis mainly aims to examine how mothers, fathers and their adolescent children influence each other’s propensities to vote for a particular party, the main explanatory variables are the propensities to vote in Wave 1. I moreover controlled for three basic variables that have been found to influence voting behaviour: gender, SES and basic ideological predispositions. These are principally stable attributes but they can be related to changes in voting propensities. I mainly included these basic attributes because of the specific and dynamic political party landscape in Belgium (Deschouwer, 2009). Many substantial movements have taken place over the past years, with the successful emergence of many new parties. For some of these new parties (e.g. N-VA), structural characteristics such as gender and educational status have proved to be important to explain structural changes in electoral support (Boonen and Hooghe, 2014). These stable attributes can play a role in explaining party preferences at one point in time, but also in explaining changes in party preferences over time.
Gender (female = 1) was only controlled for in the adolescent group, as the parent samples were subdivided into mothers and fathers. For parental SES, I used educational level, ranging from ‘elementary school’ (1) to ‘university education’ (5), and employment, which indicated whether the parent was professionally employed (1) or not (0). For the adolescents, I used educational goal, an expression of the educational level they expect to reach in the future, ranging from ‘I will probably not finish high school’ (1) to ‘university education’ (4). This variable was only measured in the second wave.
To include a basic variable for ideology, I used a variable that tapped basic left–right identification, ranging from 0 (left) to 10 (right). Additionally, I included a variable that indicated change in left–right orientation over time. The thinking behind this change in ideological predisposition might affect voting propensities (the dependent variable) as well. I calculated this by subtracting the left–right position of the respondents in Wave 2 from their left–right position in Wave 1. As such, a positive score indicates a move to the left, a negative score a move to the right, and a null-score indicates no change between both time points. A descriptive analysis shows that 56.3% of the adolescents, 52.2% of the mothers and 52.9% of the fathers somewhat changed their left–right position (mostly by 1 or 2 points) between Waves 1 and 2. A full description is included in the variable overview in online Appendix 1.
In the structural equation model (Figure 1), I regress the voting propensities of the mothers, fathers and adolescent children on their own voting intentions in Wave 2 (stability paths), and on the voting intentions of each of the other family members (cross-lagged effects). 4 The full results are presented in Table 2 and summarized schematically in Figure 2.
Structural equation model predicting propensities to vote for the Christian Democrats, Green Party, Flemish Nationalist Party and radical right party in Flanders.
Ptv: propensity to vote’: T1: Wave 1 (2012); T2: Wave 2 (2013); CFI: Comparative Fit Index; TLI: Tucker-Lewis Index; RMSEA: Root Mean Square Error of Approximation.
Note: Reported measures were obtained through a structural equation model analysis in Mplus7 statistical software package with Maximum Likelihood estimation (outcome variables are 0–10 scales). I did not use listwise deletion for the models in order to make use of all the available information. This resulted in a higher N, as I was also able to use data from incomplete father–mother–child panel triads. The results were unstandardized coefficients (B), standard errors (SE) and P-values (*p<.05, **p<.01, *** p<.001).

Schematic summary of the main results (cross-lagged effects).
First, these results indicate that there is indeed a significant positive influence from both parents on their adolescent children (H1), but we do not observe differences between mothers and fathers when it comes to the top-down influence of their adolescent children on party preferences (H2). The cross-lagged effects of both mothers and fathers who influence their child clearly go in the same direction for the four analysed parties.
In an additional test, I estimated a multi-group model to control for possible gender-specific effects (Nieuwbeerta and Wittebrood, 1995). I did observe a gender-specific pattern of influence for the Christian Democrats, where mothers influenced their daughters more strongly and fathers their sons, but this was not true for the other parties so this relation should not be generalized.
Second, looking at the horizontal influence that both adult partners have on one another, I observed a significant mutual influence, but I could not observe any patterns where one partner was more influential than the other. Only for the radical right Vlaams Belang did there seem to be a discrepancy, with mothers having a stronger influence on fathers than vice versa. Combining these first two findings, the results of this structural equation model indicate that there is no reason to assume that fathers or mothers structurally differ in their party political influence within the household, either when it comes to horizontal influence (adult partners) or top-down influence (parent–child).
Third, as hypothesized (H3), the adolescent children in the survey do not exert a clear or equally strong trickle-up influence on the voting intentions of their parents. I do, however, find a cross-lagged effect of adolescents influencing their parents’ propensity to vote for the Christian Democrats and the Green party (only mothers). These effects are small and less stable, however, and should not be generalized.
Finally, I did find differences between parties, with the radical right party the clearest exception. For this party, family members’ influence on each other was limited when compared to other parties, which could possibly be explained by a social desirability bias that also exists within family discussions.
Looking at the other variables in the model, another clear trend is that the stability of voting propensities is strong for all family members, but it is clearly stronger for the adult family members (mother and father) than for the adolescent children (stability paths in Table 2). This indicates that, for the adolescents, there is still more change and development in political preferences – and therefore more room for external influence. In this respect, these results are clearly in line with political learning theory, which underscores the important evolutions in political preferences that take place during adolescence and early adulthood (Flanagan, 2013). The strong stability coefficient indicates that political preferences, also for parties, are already present at an early stage in life, but the difference with the stability coefficients of the adult respondents indicates that there is still a margin for evolution and further crystallization of these preferences.
For ideological orientation, one of the main control variables, I saw the clear effects of both ideology at T1 and change in ideology over time. Changes in ideological orientation explained changes in propensity to vote for the leftist Green party, rightist Flemish Nationalist party and the radical right party for both adults and children, indicating a stable relationship between both orientations. Change in left–right position did not at all influence changes in propensity to vote for the centrist Christian Democratic Party. As this is a centre-right party with no strong, outspoken rightist profile, this is in line with what would be expected.
Discussion
The results of this study taught us more about the different political roles that mothers and fathers take on in the household. Fathers are more interested in politics, follow the news more frequently and have a higher level of political efficacy than their female partners. In this respect, they take on the strongest political role within the family compared to their partners and children, as H1 predicted. Interestingly, not that much seems to have changed in the division of political labour between fathers and mothers in the past 50 years. The tendencies I observed in the descriptive results were in line with the general picture that Jennings and Niemi painted using similar data from five decades ago.
My main objective was to investigate the extent to which mothers and fathers politically influence their children. Contrary to what I expected in light of some of the earliest socialization studies, the few previous research efforts in this area mostly showed a pattern of maternal dominance, which I followed in H2. In my longitudinal tests on changes in voting propensities, I did not find any evidence of a clear pattern of maternal or paternal dominance. The results from the structural equation model strongly pointed in the direction of a process of shared influence in which mothers and fathers equally contribute to the development of preferences for political parties among their adolescent children. Between (marital) partners, we seemingly observed a process of mutual influence, although these findings should be interpreted with a measure of caution as this could also be the result of a shared external influence. Couples also share the same social networks and communication environments which can influence them in the same way at the same time, without them influencing each other. This is an important alternative mechanism to keep in mind when interpreting these findings. Finally, as expected (H3), the 15- to 16-year-old adolescents in the sample did not exert a stable direct influence on the voting propensities of their parents.
These two main findings offer several lessons about the political roles that parents take on within the household. The common assumption that politics is a ‘man’s world’ is partly supported by my data, as I do find that fathers are more engaged in conventional politics. This stronger involvement in politics, however, does not give them a ‘competitive advantage’ when it comes to causing their children to absorb their preferences for political parties. The fact that the political influence of mothers and fathers is comparably strong does not mean that other dynamics do not take place. Adolescents are found to discuss party politics more often with their fathers (Hooghe and Boonen, 2015), which may lead to different socialization dynamics, for instance. The transmission from fathers to children might be the effect of political discussions, while the transmission from mothers to children might be related to a stronger emotional attachment, as was suggested in earlier socialization studies. Further disentangling these differences might prove a fruitful avenue for further research.
Use of the 0–10 propensity to vote scale also raises additional questions around the influence of parents in the development of negative evaluations of political parties, for instance. In this study, I focused on party support, but this measure could also be used to study the development of negative evaluations in a similar model, which might offer additional insights into the social identity approach to partisanship.
Finally, I started this article by underscoring the importance of testing the traditional socialization mechanisms studied in the 1960s and 1970s within the strong partisan system of the United States in a current, de-aligned democracy. In this respect, these results can teach us a lot about the applicability of these traditional transmission mechanisms in this new context. From the Belgian analyses, we learned that despite the ongoing trend towards individualized voting behaviour, it remains essential to take the immediate social network into account. This is even true for one of the most fragmented multi-party settings in Western Europe, and thus in a political system in which citizens are not born into a ‘Liberal’ or ‘Conservative’ family. The partisan cues that family members send out to one another may be weakening, and they will be less clearly observable in a multi-party setting, but this does not mean that parental influence has become less important. The young citizens in this study are not directly comparable to those Jennings and Niemi studied in the 1960s and 1970s in the strongly partisan US system. Nevertheless, they also rely strongly on parental cues, for instance, to make sense of the complex political party system that they have to negotiate.
In light of these observations, I argue that parental influence patterns should not merely be interpreted as the classical process of partisan inheritance. Within this specific European context, I would instead suggest they be interpreted as a cognitive heuristic that adolescents apply to orient themselves in a complex multi-party setting.
Footnotes
Acknowledgements
I am grateful for the helpful suggestions of Marian Sawer, co-editor of International Political Science Review, and the anonymous reviewers.
Funding
This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.
Notes
Author biography
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
