04-parent.qmd

# Parent--Child Political Interest Transmission: Do Moms Influence Their Daughters and Dads Influence Their Sons? {#sec-chap4}

Parents play a prominent role in raising their children and therefore are among the main actors who can transmit political interest to them. Political scientists have long studied the extent of parents' role in that regard and the mechanisms through which political interest can be transmitted. @campbell1960 already suggested that "interest in politics, like partisanship, is readily transmitted within the family from generation to generation" (413). Indeed, recent research has highlighted inequalities in the extent to which boys and girls match their parents' level of political interest. This chapter aims to uncover the complex role gender can play in mediating parent--child political interest transmission. When and how does political interest transmission occur? To what extent do mothers influence political interest development in their children? Do they have a greater influence on their sons or daughters? What about fathers? Do these results hold for different political topics, when political interest is measured by sector? The chapter relies on social learning theory and seeks to test this dissertation's second hypothesis highlighted in @sec-chap1: *Children's political interests are more related to political discussions with their same-gender parent(s) than their other-gender parent(s).*

## Parental Political Interest Transmission

The transmission of interest in politics between parents and children has typically been measured by asking them the same political interest questions and comparing their answers. Studies have generally shown a significant relationship between parents' political interest and their children's political interest [@beauregard2008; @janmaat2022; @neundorf2013; @prior2019; @shehata2019]. While @jennings2009 find no statistically significant relationship, their results show that successful parent--child political interest transmission occurs when the family environment is more politicized. @prior2019 estimates moderately strong Pearson correlation coefficients of parent--child political interest scores --- 0.33 in Switzerland, 0.4 in the UK, and 0.41 in Germany. By comparison, correlation coefficients of different indicators of political interest for the same individual vary between 0.6 and 0.7.

Parent--child political interest correlations vary by age. @prior2019 finds very weak (0.05) correlations at age 11, followed by a steady growth until age 15. Parent--child correlations then remain stronger when both parents share a similar level of political interest, when parental political interest is stable through time, and when children move out late from their parents' place. Children who move out early of their parents' place tend to see a quick drop in the extent to which their political interest matches their parents'. @janmaat2022 also find that, between ages 11 and 15, the gap in political interest between children whose parents are not interested in politics and children whose parents are interested in politics grows every year. @shehata2019, for their part, find the strength of the relationship between parents' news media use and children's political interest remains stable between ages 13 and 18.

While parents' political interest often seems to match their children's political interest, most studies lack the kind of data needed to establish a causal link between both. However, @prior2019 uses panel data collected among parents and children and finds a weak but noticeable causal link in parental transmission of political interest. Just like @dostiegoulet2009en, he finds that an increase in mothers' political interest is often accompanied by an increase in their children's political interest. The same goes for fathers as well as for decreases rather than increases. These trends are clearer in the United Kingdom and Germany, despite weaker evidence in Switzerland. In the first two countries, it seems reasonable to assume that a change in parents' political interest could cause a similar change in their children's political interest [@prior2019].

The main causal mechanism for this transmission process seems to be parent--child political discussions. @campbellwolbrecht2006 find children are much more likely to have political discussions at home than with peers or teachers. Scholars have long suggested that the development of political interest can happen through increasingly complex discussions about political topics between parents and children at home [@easton1969; @greenstein1965]. These discussions can be initiated either by the parent or by the child, and the more they occur, the likelier the child is to be interested in politics. This relationship between the frequency of political discussions with parents and the child's interest in politics is statistically significant in the United States [@shehata2019], China, Mexico, Japan [@mayer2004], and Canada [@dostiegoulet2009en]. Similarly, in Poland, @furman2022 find that an aggregate of discussions with parents and parents' political interest is significantly related to children's political interest, controlling for other factors, a finding that is confirmed with in-depth qualitative interview data, while @levinsen2015 find the same for political discussions with young people's best friends in Denmark. Furthermore, @shehata2019 find that changes in the frequency of political discussions with parents positively predict changes in adolescents' level of political interest.

It is also important to recognize that children can initiate political discussions with their parents --- and potentially shape their parents' interest in politics. @mcdevitt2002 find evidence of trickle-up political socialization, in which adolescent children initiate political discussions with parents, who react by increasing their news consumption or finding other ways to gain knowledge about politics in order to maintain a leadership role in the family. In this study, children's interest is first triggered through a civics curriculum. @york2019 even finds that adolescents' news use and political discussions with peers have a positive influence on their political discussions with parents, while the opposite effects --- political discussions with parents influencing adolescents' news use or political discussions with peers --- are null. In a similar vein, @stattin2022 show that adolescents' initial level of political interest can predict changes in their perceptions of their parents' political interest, while their parents' initial political interest (as perceived by the child) does not predict changes in their own future political interest.

Yet, most other studies about the relative influence of various socialization agents on the development of political interest in children and adolescents mostly find that parents play a larger role in transmitting political interest than any other socialization agent --- including friends, media and schools (@dostiegoulet2009en and @shehata2019 but not @koskimaa2015, who find a larger effect for friends). Trickle-down political socialization --- from parents to children --- therefore remains an important explanation of children's development of political interest.

## Gender Differences in Parental Transmission

Research has found that the trickle-down effect of political interest from parents to children works in gendered ways, and much of that evidence suggests transmission is stronger for parent--child pairs of the same gender. From the perspective of parents, mothers' political interest has a stronger effect on their daughters than sons' political interest, while fathers' political interest has a stronger effect on their sons' political interest [@beauregard2008; @owen1988; @prior2019]. Disregarding parents' gender, @sabella2004political finds that parents' potential to transmit political interest to their sons is stronger than to their daughters. From the perspective of children, daughters' political interest seems to be influenced mostly by their mothers, with the mother--daughter political interest link being stronger than all other combinations [@beauregard2008; @owen1988; @prior2019], although @rebenstorf2004political strikes a discordant note, finding that fathers' potential to transmit political interest to their children is stronger than mothers'. Finally, it is not clear whether sons' political interest is influenced mostly by their father [@beauregard2008; @owen1988] or mother [@prior2019].

@prior2019 studies the strength of parent--child political interest correlations by gender using more recent panel data from the United Kingdom, Germany and Switzerland. In the UK dataset that he analyses, @prior2019 finds that mother--daughter political interest correlations are the strongest (0.43), while mother--son, father--daughter and father--son correlations all sit between 0.31 and 0.34. In his German dataset, all four pairs are relatively close together, but exact numbers are not provided by the author. Finally, in the Swiss dataset, mother--daughter correlations are the strongest (0.36), followed by mother--son (0.31), father--son (0.21) and father--daughter (0.2). There is overall a stronger causal effect of mothers' political interest on their children's political interest, compared with fathers.^[@kestilakekkonen2023 also find stronger effects of mothers on children when it comes to the transmission of political self-efficacy, but do not find a stronger relationship between parents and children of the same gender.] Using data collected among teenagers in Germany, @oswald1998 find that the gap in the frequency of political discussions with mothers and fathers remains stable between ages 16 and 18.

Studies have investigated the gender patterns in parent--child political discussions, but the amount of political discussions does not seem to vary based on parents' and children's gender. While earlier studies found that children discuss politics more often with their fathers than mothers [@levinsen2015; @noller1985; @oswald1998], most of the recent research has found no significant difference between fathers and mothers [@hooghe2015; @mayer2004; @shulman2014]. @noller1985 also find that parents discuss politics more often with their sons. However, most studies again find a different pattern in which parents discuss politics with their daughters as much as with their sons [@dowse1971; @lawless2015; @mayer2004]. Overall, it seems reasonable to assume, as @hooghe2015 do, that political discussions involving the father tend to revolve mostly around partisan politics, while political discussions with mothers might center on other topics --- presumably health care, education, gender issues, and so on.

The nature of discussions about politics could also vary by gender, with more conflictual discussions happening between sons and fathers, or between mothers and daughters, than between other-gender pairs. @vuchinich1987 shows the opposite to be true: parents are more likely to begin conflictual interactions with their other-gender children than same-gender children, while children initiate conflictual interactions with their mother more than with their father. It is not clear whether these patterns also apply to political discussions.

While past studies by @mayer2004 and others have found no differences in the frequency of political discussions by child gender or parent gender, these studies have not studied whether the specific topics of political discussions in the family could vary by parent gender or child gender. Are fathers more likely than mothers to discuss agency-related political topics with their children? Are parents more likely to discuss cooperation-focused political topics with their daughters than sons? Hypotheses 2a--2d provide hypotheses about the frequency of political discussions by gender in the family unit.

- *__Hypothesis 2a__: Mothers are more likely than fathers to discuss the politics of health care and education with their children.*

- *__Hypothesis 2b__: Fathers are more likely than mothers to discuss law and crime, international affairs, and partisan politics with their children.*

- *__Hypothesis 2c__: Parents are more likely to discuss the politics of health care and education with their daughters than sons.*

- *__Hypothesis 2d__: Parents are more likely to discuss law and crime, international affairs, and partisan politics with their sons than daughters.*

Hypotheses 2a and 2b focus their attention on discrepancies between mothers' and fathers' likelihood of discussing certain topics with their children, while Hypotheses 2c and 2d focus their attention on discrepancies between daughters and sons in their likelihood of discussing politics with --- or being told about politics by --- their parents. Hypotheses 2a and 2c suggest that parent gender is a significant factor that explains which topics will be discussed with children, and Hypotheses 2b and 2d suggest that child gender is a key variable that explains which topics they will discuss with their parents.

Parents' education can also affect their children's interest in politics. Status transmission theory suggests that "well-educated parents are more likely to provide a politically stimulating home environment" and therefore have politically engaged children [@gidengil2016b, 373]. It is not clear if status transmission theory works in gendered ways. @beauregard2008 and @janmaat2022 find that parents' education is positively associated with their adolescents' political interest regardless of parents' gender, while other studies find a stronger effect of one parent. @neundorf2013 and @koskimaa2015 find a positive relationship between fathers' education and children's interest in politics, @sanjuan2022 instead find a positive relationship for mothers, @jennings2009 find no relationship for both fathers and mothers, and @koskimaa2015 even find a *negative* relationship between mothers' education and children's political interest.

Other parental characteristics can also influence the development of children's political interest in gendered or non-gendered ways. @gidengil2010 and @cicognani2012 find that a mother's level of political participation has a positive link with her daughter's political interest. @jennings2009 do not find a relationship between parents' income and children's political interest, but @neundorf2013 find a positive relationship between fathers' income and children's political interest, which they suggest may be due to an indirect process in which fathers' income increases children's level of education and development of higher class civic attitudes, which then increases their level of interest in politics. @borkowska2024 find that political interest transmission patterns vary between immigrant and non-immigrant families, with intergenerational transmission somewhat weaker in immigrant families than other families, notably because other factors --- such as socialization in a good-performing democracy or not --- have a more significant influence. The authors find no moderating effect for naturalization.

## Social Learning Theory

As highlighted in @sec-chap1, social learning theory suggests that children learn by observing their parents' behaviour, attitudes, habits and values, and model their behaviour, attitudes, habits and values after them [@gidengil2016b; @shehata2019]. This process applies to the transmission of political interest [@jennings2009; @prior2019; @shehata2019]. @prior2019 suggests the transmission process starts in early adolescence when children still live in the family home but start acquiring an understanding of what politics is. Its effectiveness is higher when parents give consistent and strong cues about their own level of political interest [@jennings2009; @prior2019]. Parents' political opinions and leanings are also more likely to be known by their children when the family environment is politicized since this environment can foster social learning [@jennings2009; @neundorf2013]. In a political home environment, children can feel social pressure to become interested in politics to create or maintain a sense of social belonging in the family unit, be exposed to more news media content, listen to or participate in more political discussions at home, or all of these [@shehata2019]. @bandura1969 further explains that observer--model similarity leads children to model their behaviour, values, attitudes and habits on models that resemble them, notably their parent(s) of the same gender.

Social learning theory and observer--model similarity both assume that political interest should be more strongly correlated between mothers and daughters and between fathers and sons than any other combination. While this assumption has somewhat effectively been tested in past research, mothers seem to have an overall stronger influence on the development of political interest in their children. Moreover, it remains unclear if this finding applies similarly across various political topics, or if the cooperation-oriented topics in which women typically report being more interested --- health care and education --- can better be transmitted by mothers while those assertion-focused topics in which men typically report being more interested --- law and crime, international affairs, and partisan politics --- can better be transmitted by fathers. Four hypotheses are therefore tested alongside each other:

- *__Hypothesis 2e__: Children's political interests are more affected by political discussions with their same-gender parent(s) than their other-gender parent(s).*

According to Hypothesis 2e and social learning theory more generally, a parent's interest in a specific political topic should influence interest in that topic more strongly for their same-gender children than other-gender children. On average, a mother is expected to have more transmission potential of her interest in health care issues to her daughters than sons through political discussion, for example. Given the importance of parents' role in socialization in childhood and adolescence, this transmission process would explain part of the gender gap in interest in topics such as health care and partisan politics that exists among adults.^[Hypothesis 2e supposes similar effects of parental discussions for different topics and suggests the key factor explaining interest transmission is parent--child gender congruence, but there are other possible answers to the question of how political interest gender gaps could be transmitted by parents, including the nature of the topics themselves. Mothers may have more transmission potential of interest in cooperation-focused topics for which the average woman reports more interest, and fathers may have more transmission potential of interest in assertion-focused topics for which the average man reports more interest. Since @campbellrosie2008 suggest that women tend to see political issues through the lens of cooperation while men tend to see them through a lens of self-assertion, it could be the case that parental socialization of boys favours assertion-focused political topics while parental socialization of girls favours cooperation-focused political topics.] It is also possible that mothers have more transmission potential than fathers for interest in all topics, a finding that would be consistent with @prior2019's finding that the mother--daughter transmission link is the strongest.

- *__Hypothesis 2f__: Children's political interests become more and more affected by political discussions with their parent(s) as they age.*

Finally, Hypothesis 2f tests the time trends in parental influence over children's political interest, testing results put forward by @janmaat2022, @prior2019 and @shehata2019, but relying on measures of parent--child political discussions rather than parental political interest. The hypothesis implies that social learning and observer--model similarity keep affecting children more and more as they age, with girls developing more interest in topics they discuss with their mothers and boys developing more interest in topics they discuss with their fathers.^[For the sake of simplicity, the effect is presumed to be more or less constant across political topics and all parent--child pairs --- mother--son, mother--daughter, father--son, father--daughter.]

## Data and Methods

The 2022--23 Children Political Interest Survey (CPIS) is used to study relationships between students' interest in certain topics and parents' discussions of these same topics. This web-collected bilingual dataset includes survey responses from 698 Canadian children and adolescents aged 9 to 18. The CPIS includes information about students' interest in five political topics: health care, international affairs, law and crime, education, and partisan politics. Further information about the dataset and question wording for interest questions are found in @sec-chap2.

In order to determine the importance of parents depending on their gender, the following question was asked to children: "Which parent do you discuss most often with? (a) Mother; (b) Father; (c) Both equally".  Several questions are then used to assess the role of parents in transmitting interest to their children. Given the important role of political discussion in the transmission of political interest, this variable is used to measure parents' transmission potential: "For each of the following topics, which parent do you discuss most often with?" Answers are either "My mother", "My father" or "Don't know/Prefer not to answer".^[Students who do not have one parent of either gender are removed from the analysis.] Second, students are asked "Among these five topics, which one do you discuss most often with your mother(s)?" Each of the five topics is listed, and the same question is then asked about the father(s). While topics such as health care and education can be spoken about without referring to their political aspects, children were given examples of political issues related to each of these topics shortly before in the same survey when they were asked which topic they were most interested in. These questions are limited by the fact that data was collected among children only --- the perceptions of their parents may differ. Yet, the relative levels of discussions of these topics relative to one another and between mothers and fathers should not theoretically be affected by the study's design.

All multilevel regression results presented in this chapter include classroom fixed effects. A simple multilevel regression model with one explanatory variable --- often a given socialization agent's interest in a specific topic --- is followed by multiple regression models with control variables for two blocs of variables: (1) socio-economic status variables, (2) all other socialization agents than parents (peers, teachers, and influencers) and a squared term for age. Socio-economic status variables include confounding variables that have been linked with political interest and could therefore mediate any given relationship found between two persons' political interest. These include gender, age, language, ethnicity, and immigrant status.

## Results

### Parent--Child Political Discussions by Gender

@fig-parents shows, for each topic, which parent students report discussing the most often with. Out of 698 students, including missing data and non-responses, 61% of students report discussing health care more often with their mother and 13% with their father. Similarly, 55% say they discuss education with their mother more than with their father, while 19% say they discuss it more with their father. On the contrary, 43% of students say they discuss law and crime more often with their father while 25% discuss this topic more often with their mother; 36% discuss partisan politics more often with their father while 17% discuss it more often with their mother; and 47% discuss international affairs more often with their father while 19% discuss it more with their mother. There are more non-answers and missing data for partisan politics, which presumably means no parent discusses the topic at home with their children. When these results are broken down by students' gender ("All discussions" panel on the left), very similar results are found for boys and girls, with one exception: in general, for all types of discussions --- regardless of their political nature --- 24% of boys say they have more discussions with their mother, compared with 43% of girls who report the same. Boys are more likely not to respond to that question, which suggests they may have similar amounts of discussions with both parents.

```{r}
#| label: fig-parents
#| echo: false
#| fig-cap: "Topic Most Often Discussed with Parents by Child Gender, 2022 CPIS data"
knitr::include_graphics("_graphs/ParentTopicsGrey.pdf")
```

One could argue that children discuss the politics of health care and education with their mothers simply because they spend more time discussing with their mothers than fathers in general. This indeed seems to be the case for girls --- 82% of them discuss mostly with their mothers in general, which is more than the 79% who discuss education mostly with their mothers, and only slightly less than the 84% who discuss health care mostly with their mothers. For boys however, discussions with their mothers about the politics of health care (81%) and education (70%) are more common than discussions overall (62%).

These results match with what previous literature has found about women reporting more interest in education and health care, and men reporting more interest in partisan politics, law and crime, and international affairs. Hypotheses 2a and 2b are therefore corroborated. Also in line with the literature, parents seem to talk about various political topics just as much with their sons as with their daughters. Hypotheses 2c and 2d, which hypothesized agency-related topics would be more prevalent in discussions with sons while cooperation-focused topics would be more prevalent in discussions with daughters, are not corroborated.

@fig-parentsmomdad shows the extent to which mothers discuss each of the five topics and the same for fathers. Mothers overwhelmingly discuss education and health care according to their children, while fathers discuss international affairs more than other topics but seem like a more heterogeneous group. Despite partisan politics being the least discussed topic by both parents, fathers are much more likely to discuss it than mothers according to this metric. For law and crime too, fathers are more likely to discuss it. Again, when these results are broken down by students' gender, very similar results are found for boys and girls.

```{r}
#| label: fig-parentsmomdad
#| echo: false
#| fig-cap: "Topic Most Often Discussed by Mothers and Fathers, 2022 CPIS"
knitr::include_graphics("_graphs/ParentTopicsMomDadGrey.pdf")
```

The gender gap in political topics discussed is pretty stark between mothers and fathers, according to their children's assessments. It is important to specify that given the question's phrasing, it is not clear if these interactions are initiated by parents or by their children, and if it is reflective of top-down or trickle-up political socialization [@mcdevitt2002]. Moreover, the question does not ask children to specify a percentage of interactions about each topic started by the mother or father; it simply asks them to pick the parent most likely to have these discussions with them. There is no assumption that mothers are the only ones talking about health care and fathers the only ones to talk about international affairs, but on average, they are more likely to raise this topic than the other-gender parent --- with the caveat that children may have been thinking about health care and education in a broader sense than simply their political aspects, despite prior prompting that should encourage them to think about political aspects. Women remain over-represented in the care sector and are more likely to care for children and relatives at home in Canada, which could explain these results.

Regardless, the fact that results among boys and girls both strongly point in the same direction is revealing. It makes little doubt that mothers and fathers speak differently about politics when they raise the topic with their children. Both questions were formulated in concrete ways, asking children the extent to which they discuss these topics with their parents. There is no Hawthorne effect among kids; they would not have an expected answer. Parents speak in starkly different ways about politics to their children, and this confirms that political socialization is a deeply gendered process.

### Topics Parents Discuss the Most

When it comes to parents' role in political interest transmission, Table \ref{tab:lmeParentSESInterac} shows the relationship between students' interest in each of the five topics and the gender of the parent who discusses the topic with them the most. The top part of the table shows determinants of interest among boys, while the bottom part does the same for girls. Column "All" is a stacked dataset that aggregates all parent--child linkages matched by topic. Children's interest in either of the five topics does not seem to be related to the gender of the parent who discusses the topic the most; the topic-by-topic analysis shows all relationships are statistically insignificant. However, in stacked data, the fact that their father discusses a topic more than their mother increases sons' interest in that topic by an average of 0.7 points on an 11-point scale (*p*<0.001). This effect size represents more than half of the 0.9-point gender gap in interest in partisan politics and 1-point gender gap in interest in international affairs seen in @sec-chap3. The gender of the parent who discusses a topic more does not have a significant effect on daughters' interest in that topic, a finding that contradicts theoretical expectations that mothers play a particularly important role in their daughters' political socialization.^[Table \ref{tab:lmeParent} in @sec-appendix8 shows very similar findings when control variables are removed.] No variables among those tested significantly predict girls' political interest. Table \ref{tab:lmeParentBoysGirls} in @sec-appendix7 replicates the analysis while putting boys and girls in the same model. A significant positive effect of fathers on their children's political interest is found, supplemented by a negative term associated with being a white girl.

\begin{table}
\centering\centering
\caption{Interest in Topic by Gender of Parent who Discusses that Topic the Most (With Interactions) \label{tab:lmeParentSESInterac}}
\centering
\fontsize{6}{8}\selectfont
\begin{tabular}[t]{lcccccc}
\toprule
  & All & Health care & International affairs & Law and crime & Education & Partisan politics\\
\midrule
\addlinespace[0.5em]
\multicolumn{7}{l}{\textit{Boys}}\\
\midrule \hspace{1em}(Intercept) & 8.760 & 10.988 & -3.757 & 10.621 & 31.363* & 37.670+\\
\hspace{1em} & (8.739) & (15.860) & (16.316) & (16.266) & (14.909) & (19.217)\\
\hspace{1em}Mother discusses topic more than father & -0.651*** & -0.388 & -0.380 & -0.345 & 0.232 & -0.088\\
\hspace{1em} & (0.161) & (0.399) & (0.397) & (0.390) & (0.388) & (0.487)\\
\hspace{1em}Age & -0.649 & -1.179 & 1.447 & -0.867 & -4.016+ & -4.790+\\
\hspace{1em} & (1.198) & (2.189) & (2.240) & (2.261) & (2.053) & (2.657)\\
\hspace{1em}Age squared & 0.027 & 0.049 & -0.050 & 0.035 & 0.146* & 0.168+\\
\hspace{1em} & (0.041) & (0.075) & (0.076) & (0.078) & (0.070) & (0.091)\\
\hspace{1em}Ethnicity (1 = white) & 0.316+ & 0.124 & 1.082** & -0.124 & -0.152 & 0.517\\
\hspace{1em} & (0.192) & (0.373) & (0.385) & (0.406) & (0.400) & (0.510)\\
\hspace{1em}Immigrant & -0.215 & 0.228 & -1.025* & 0.253 & -0.420 & -0.791\\
\hspace{1em} & (0.259) & (0.499) & (0.519) & (0.542) & (0.532) & (0.714)\\
\hspace{1em}English spoken at home & -0.406 & -0.334 & -1.252 & -0.429 & -0.376 & -0.144\\
\hspace{1em} & (0.406) & (0.777) & (0.803) & (0.829) & (0.818) & (1.132)\\
\hspace{1em}French spoken at home & -0.017 & 0.149 & -0.882+ & 0.040 & 0.046 & 0.343\\
\hspace{1em} & (0.230) & (0.444) & (0.453) & (0.480) & (0.470) & (0.603)\\
\hspace{1em}SD (Intercept Class) & 0.757 & 0.657 & 0.829 & 0.385 & 0.394 & 0.001\\
\hspace{1em}SD (Observations) & 2.642 & 2.339 & 2.413 & 2.629 & 2.694 & 2.977\\
\hspace{1em}Num.Obs. & 1107 & 233 & 220 & 222 & 246 & 186\\
\hspace{1em}R2 Marg. & 0.021 & 0.029 & 0.067 & 0.015 & 0.041 & 0.049\\
\addlinespace[0.5em]
\multicolumn{7}{l}{\textit{Girls}}\\
\midrule \hspace{1em}(Intercept) & 12.388+ & 10.465 & 25.279+ & 11.236 & 17.735 & 44.466*\\
\hspace{1em} & (7.185) & (11.106) & (13.521) & (13.799) & (11.679) & (19.529)\\
\hspace{1em}Mother discusses topic more than father & -0.232 & 0.194 & -0.009 & -0.019 & -0.383 & -0.123\\
\hspace{1em} & (0.176) & (0.450) & (0.431) & (0.416) & (0.449) & (0.481)\\
\hspace{1em}Age & -1.193 & -1.144 & -2.961 & -0.815 & -2.112 & -5.743*\\
\hspace{1em} & (0.997) & (1.552) & (1.892) & (1.930) & (1.640) & (2.695)\\
\hspace{1em}Age squared & 0.046 & 0.048 & 0.106 & 0.033 & 0.081 & 0.198*\\
\hspace{1em} & (0.034) & (0.054) & (0.066) & (0.067) & (0.057) & (0.092)\\
\hspace{1em}Ethnicity (1 = white) & -0.346 & -0.287 & 0.224 & -0.227 & -0.453 & -0.124\\
\hspace{1em} & (0.219) & (0.399) & (0.483) & (0.468) & (0.425) & (0.566)\\
\hspace{1em}Immigrant & -0.143 & 0.234 & -0.548 & -1.024 & 0.701 & -0.139\\
\hspace{1em} & (0.287) & (0.522) & (0.648) & (0.657) & (0.557) & (0.680)\\
\hspace{1em}English spoken at home & -0.471 & -0.106 & -0.462 & -1.111 & 0.195 & 0.046\\
\hspace{1em} & (0.371) & (0.631) & (0.800) & (0.765) & (0.726) & (0.854)\\
\hspace{1em}French spoken at home & 0.083 & 0.219 & -0.127 & -0.703 & 0.332 & 0.509\\
\hspace{1em} & (0.241) & (0.440) & (0.525) & (0.527) & (0.473) & (0.611)\\
\hspace{1em}SD (Intercept Class) & 0.657 & 0.604 & 0.001 & 0.656 & 0.679 & 0.170\\
\hspace{1em}SD (Observations) & 2.684 & 2.380 & 2.739 & 2.787 & 2.595 & 2.716\\
\hspace{1em}Num.Obs. & 992 & 226 & 192 & 203 & 221 & 150\\
\hspace{1em}R2 Marg. & 0.017 & 0.035 & 0.020 & 0.029 & 0.054 & 0.038\\
\bottomrule
\multicolumn{7}{l}{\rule{0pt}{1em}+ p $<$ 0.1, * p $<$ 0.05, ** p $<$ 0.01, *** p $<$ 0.001}\\
\multicolumn{7}{l}{\rule{0pt}{1em}Method: Multilevel linear regression}\\
\multicolumn{7}{l}{\rule{0pt}{1em}Fixed Effects: Classroom}\\
\multicolumn{7}{l}{\rule{0pt}{1em}Reference Category for Language: Other languages spoken at home}\\
\end{tabular}
\end{table}

### Topics Most Often Discussed with Mothers

Table \ref{tab:lmeAgentsSES} shows students' interest in each of the five topics depending on whether this topic is the one they most often discuss with various socialization agents, starting with mothers. The top part of the table shows determinants of interest among boys, while the bottom part does the same for girls. Again, column "All" is a stacked dataset that aggregates all socialization agent--child linkages matched by topic. For boys, interest in either of the five topics or the stacked data is unrelated to their discussion of these topics with their mothers. For girls, interest in education is related to their mothers discussing education (*p*<0.01). For other topics, no statistically significant effect is found^[The coefficient for partisan politics was dropped from the model for rank deficiency.] --- but all coefficients are still positive. In the stacked data, if a girl's mother discusses one of the five topics most, her interest in that topic is expected to increase by 0.7 points on the 11-point scale (*p*<0.05), almost as large as the gender gaps in interest in international affairs (1) or partisan politics (0.9).

\begin{table}
\centering\centering
\caption{Interest in Topic Most Often Discussed with Socialization Agents \label{tab:lmeAgentsSES}}
\centering
\fontsize{6}{8}\selectfont
\begin{tabular}[t]{lcccccc}
\toprule
  & All & Health care & International affairs & Law and crime & Education & Partisan politics\\
\midrule
\addlinespace[0.5em]
\multicolumn{7}{l}{\textit{Boys}}\\
\midrule \hspace{1em}(Intercept) & 6.011 & -17.012 & -15.964 & 31.164 & 33.365 & 8.199\\
\hspace{1em} & (12.054) & (24.194) & (17.744) & (20.788) & (22.838) & (23.231)\\
\hspace{1em}Topic most discussed with mother? & -0.279 & 0.997+ & -0.243 & 0.021 & -0.729 & 0.401\\
\hspace{1em} & (0.288) & (0.558) & (0.815) & (1.338) & (0.600) & (1.612)\\
\hspace{1em}Topic most discussed with father? & 0.768** & 1.431 & 0.310 & 1.291* & -0.420 & 2.111+\\
\hspace{1em} & (0.292) & (1.313) & (0.458) & (0.607) & (0.764) & (1.057)\\
\hspace{1em}Topic most discussed with female friends? & 0.664* & -0.433 & -0.647 & 2.471** & 1.797** & 1.122\\
\hspace{1em} & (0.298) & (0.766) & (0.489) & (0.857) & (0.606) & (2.141)\\
\hspace{1em}Topic most discussed with male friends? & 0.594+ & -2.723* & 0.857+ & 0.167 & -0.085 & 2.240\\
\hspace{1em} & (0.315) & (1.209) & (0.481) & (0.670) & (0.778) & (1.468)\\
\hspace{1em}Topic most discussed by teacher? & 0.326 & 1.481 & 1.133* & -0.284 & -0.443 & -0.003\\
\hspace{1em} & (0.298) & (0.967) & (0.471) & (1.201) & (0.660) & (1.359)\\
\hspace{1em}Topic most discussed by social media influencer? & 0.703* & 0.864 & 0.115 & 0.316 & 0.618 & -1.216\\
\hspace{1em} & (0.295) & (0.667) & (0.456) & (0.712) & (1.161) & (2.173)\\
\hspace{1em}Age & -0.557 & 2.737 & 2.971 & -4.208 & -4.067 & -0.873\\
\hspace{1em} & (1.691) & (3.393) & (2.497) & (2.928) & (3.196) & (3.277)\\
\hspace{1em}Age squared & 0.029 & -0.088 & -0.099 & 0.155 & 0.146 & 0.038\\
\hspace{1em} & (0.059) & (0.118) & (0.087) & (0.102) & (0.111) & (0.115)\\
\hspace{1em}Ethnicity (1 = white) & 0.087 & -0.648 & 0.454 & 0.258 & -0.369 & 0.469\\
\hspace{1em} & (0.286) & (0.629) & (0.537) & (0.560) & (0.688) & (0.681)\\
\hspace{1em}Immigrant & 0.044 & 0.655 & -1.107 & 1.579+ & -0.325 & -0.704\\
\hspace{1em} & (0.415) & (0.892) & (0.728) & (0.796) & (0.952) & (0.938)\\
\hspace{1em}English spoken at home & 0.426 & -0.127 & -0.408 & 1.192 & 1.521 & 0.307\\
\hspace{1em} & (0.750) & (1.537) & (1.216) & (1.438) & (1.599) & (1.564)\\
\hspace{1em}French spoken at home & 0.861* & 0.973 & -0.303 & 1.572* & -0.217 & 0.980\\
\hspace{1em} & (0.389) & (0.873) & (0.665) & (0.752) & (0.886) & (0.858)\\
\hspace{1em}SD (Intercept Class) & 0.940 & 1.185 & 0.000 & 1.057 & 0.182 & 0.951\\
\hspace{1em}SD (Observations) & 2.369 & 2.266 & 1.993 & 2.100 & 2.665 & 2.472\\
\hspace{1em}Num.Obs. & 458 & 90 & 92 & 92 & 92 & 92\\
\hspace{1em}R2 Marg. & 0.114 & 0.171 & 0.219 & 0.210 & 0.144 & 0.179\\
\addlinespace[0.5em]
\multicolumn{7}{l}{\textit{Girls}}\\
\midrule \hspace{1em}(Intercept) & -26.389+ & -9.169 & -38.483 & 3.836 & -23.826 & -1.653\\
\hspace{1em} & (15.858) & (23.812) & (31.457) & (27.142) & (26.186) & (28.375)\\
\hspace{1em}Topic most discussed with mother? & 0.729* & 0.259 & 1.667 & 1.769+ & 1.798** & \\
\hspace{1em} & (0.318) & (0.555) & (1.689) & (0.958) & (0.623) & \\
\hspace{1em}Topic most discussed with father? & 0.454 & 0.226 & -0.130 & 1.935* & 0.564 & -0.059\\
\hspace{1em} & (0.311) & (0.822) & (0.713) & (0.721) & (0.711) & (1.010)\\
\hspace{1em}Topic most discussed with female friends? & 0.468 & -0.060 & 0.852 & -0.401 & 0.656 & 2.236\\
\hspace{1em} & (0.320) & (0.637) & (0.856) & (0.762) & (0.611) & (2.535)\\
\hspace{1em}Topic most discussed with male friends? & 0.647* & -0.788 & -0.177 & -0.228 & 0.213 & 3.291**\\
\hspace{1em} & (0.307) & (0.833) & (0.760) & (0.607) & (0.756) & (1.112)\\
\hspace{1em}Topic most discussed by teacher? & 0.085 & -0.902 & -0.070 & -0.763 & 0.233 & 2.856***\\
\hspace{1em} & (0.308) & (1.193) & (0.785) & (0.751) & (0.602) & (0.775)\\
\hspace{1em}Topic most discussed by social media influencer? & 0.812** & 1.418* & -0.444 & 1.494* & -1.910+ & 1.737\\
\hspace{1em} & (0.309) & (0.648) & (0.703) & (0.672) & (0.977) & (1.525)\\
\hspace{1em}Age & 3.932+ & 1.299 & 5.885 & 0.279 & 3.050 & 0.623\\
\hspace{1em} & (2.146) & (3.271) & (4.311) & (3.731) & (3.583) & (3.918)\\
\hspace{1em}Age squared & -0.126+ & -0.028 & -0.195 & -0.009 & -0.086 & -0.020\\
\hspace{1em} & (0.072) & (0.112) & (0.147) & (0.128) & (0.122) & (0.134)\\
\hspace{1em}Ethnicity (1 = white) & -0.186 & 0.646 & -0.370 & -0.531 & 0.121 & -0.490\\
\hspace{1em} & (0.346) & (0.634) & (0.768) & (0.717) & (0.691) & (0.649)\\
\hspace{1em}Immigrant & -0.373 & -0.365 & -1.122 & -2.478* & 0.761 & -0.528\\
\hspace{1em} & (0.441) & (0.837) & (1.052) & (0.960) & (0.878) & (0.871)\\
\hspace{1em}English spoken at home & -0.290 & -0.401 & -0.722 & -0.549 & 0.270 & -0.282\\
\hspace{1em} & (0.487) & (0.814) & (1.027) & (0.886) & (0.922) & (0.879)\\
\hspace{1em}French spoken at home & 0.272 & -0.454 & 0.057 & -0.284 & 0.855 & -0.187\\
\hspace{1em} & (0.412) & (0.710) & (0.888) & (0.775) & (0.768) & (0.769)\\
\hspace{1em}SD (Intercept Class) & 1.069 & 0.000 & 0.442 & 0.001 & 0.872 & 0.720\\
\hspace{1em}SD (Observations) & 2.372 & 2.186 & 2.607 & 2.328 & 2.259 & 2.170\\
\hspace{1em}Num.Obs. & 387 & 78 & 77 & 76 & 78 & 78\\
\hspace{1em}R2 Marg. & 0.096 & 0.196 & 0.088 & 0.302 & 0.230 & 0.267\\
\bottomrule
\multicolumn{7}{l}{\rule{0pt}{1em}+ p $<$ 0.1, * p $<$ 0.05, ** p $<$ 0.01, *** p $<$ 0.001}\\
\multicolumn{7}{l}{\rule{0pt}{1em}Method: Multilevel linear regression}\\
\multicolumn{7}{l}{\rule{0pt}{1em}Fixed Effects: Classroom}\\
\multicolumn{7}{l}{\rule{0pt}{1em}Reference Category for Language: Other languages spoken at home}\\
\end{tabular}
\end{table}

Among significant predictors of girls' interest in all political topics, only social media influencers have a marginally larger substantial (0.8-point) effect, which highlights the important role mothers can have among broader socialization influences. Among significant predictors of boys' interest in all political topics, several other socialization agents seem to play an important role.^[@sec-chap6 briefly discusses results with regards to the influence of teachers and social media influencers.]

### Topics Most Often Discussed with Fathers

Table \ref{tab:lmeAgentsSES} then shows students' interest in a specific topic depending on it being the topic they most often discuss with their father(s). For boys, interest in law and crime is related to their fathers discussing law and crime (*p*<0.05). In the stacked data, if a boy's father discusses one of the five topics most, their interest in that topic is expected to increase by 0.8 points on an 11-point interest scale (*p*<0.01), an effect that is almost as large as the gender gap in interest in partisan politics. For girls, interest in law and crime is also related to their fathers discussing law and crime (*p*<0.05). In the stacked data, however, no significant effect is found. None of the other topic-specific father--son and father--daughter relationships are statistically significant at the 95% confidence level.^[Coefficients for peers are analyzed in @sec-chap5.]

Overall, results from Table \ref{tab:lmeAgentsSES} lend support to Hypothesis 2e, which suggests that children's interest in a given political topic should be more related to their same-gender parent discussing that topic with them than their other-gender parent discussing it with them. Fathers' influence on sons seems to be similar to mothers' influence on daughters, contrary to some past research in which the influence of mothers on political interest development was deemed to be generally stronger than fathers. Table \ref{tab:lmeParentSESInterac} also shows that when a topic is discussed by their father rather than their mother, their sons' interest in that topic is more likely to increase. For girls, whichever parent discusses a topic more does not seem to significantly predict their interest in that topic when the question is asked this way --- in particular for white girls. Notably, results do not suggest that the influence of mothers in the transmission of political interests is greater than the influence of fathers.

I conducted two tests to verify whether the relationship between parental discussion and child interest is robust to the inclusion and removal of confounding variables. First, Tables \ref{tab:lmeMother} and \ref{tab:lmeFather} in @sec-appendix8 reproduce coefficients for mothers and fathers from Table \ref{tab:lmeAgentsSES} without controls. Generally, the same relationships are found, although the influence of mothers on their daughters and of fathers on their sons increases by 1 point and becomes significant at a higher level. Father--daughter interest transmission in the stacked data becomes significant at *p*<0.01 but remains substantively smaller than both father--son and mother--daughter transmission. For all parent--child pairs, law and crime interest transmission also becomes marginally significant at least at the *p*<0.05 level. @fig-mother and @fig-father in @sec-appendix7 also include a model in which SES is controlled for but not discussions with other socialization agents. Results are similar to these models without controls. Overall, including these robustness checks, out of 48 models for boys and 47 for girls, there are 2 significant positive relationships between mothers and sons, 8 between mothers and daughters, 10 between fathers and sons, and 5 between fathers and daughters, suggesting there is more evidence of same-gender interest transmission. All other relationships are statistically insignificant.

Second, Table \ref{tab:lmeAgentsBoysGirls} in @sec-appendix7 replicates the Table \ref{tab:lmeAgentsSES} analysis while putting boys and girls in the same model. Here, the influence of fathers remains strong, positive and significant at the 95% confidence level, but the coefficient for mothers becomes *negative* --- discussion with the mother is not associated with more interest in any of the five topics or the stacked data. However, the interaction term between mother discussion and child interest is significant and even larger than the coefficient for the influence of fathers (1 vs. 0.8), again confirming mothers do play a role in the interests of their daughters.

Overall, the topic-by-topic analysis does not show a consistent pattern, with relatively few significant coefficients and a lower sample size. However, Table \ref{tab:lmeAgentsSES} shows that interest in *some* cooperation-oriented topics such as education is more likely to be transmitted by mothers --- at least for girls --- while interest in law and crime is likely to be transmitted by either parent.

### Age Trends

@fig-discussparentyoses shows relationships between parents' discussion of certain topics and children's interest in those topics broken down by age group. For the sake of simplicity and to avoid small sample sizes, topic-by-topic analyses are excluded, with all analyses using stacked data of each parent discussion--child interest pair. Results are mixed: older teenagers (ages 16--18) seem at first glance to be more influenced by their parents' discussion of certain topics than 10--15-year-olds, but gender gaps between both age groups are not significantly different. Yet, for girls, the influence of both their father and mother is only statistically significant among 16--18 year-olds, with a somewhat larger coefficient than girls aged 10--15. No difference between age groups is found among boys. However, for boys aged 16--18, when their parent most interested in a specific topic is their father, they become more interested in that topic, while this relationship is weaker and non-significant for boys aged 10--15. These results provide at best mixed evidence for Hypothesis 2f: aging could make more statistically significant relationships emerge, but the effect sizes by age group only become wider by non-significant margins. Yet, among 10--15-year-olds, none of the parent--child political interest transmission links are statistically significant, which could indicate that early parental socialization is more limited than what was previously thought. It could be the case that due to a more limited sample size, smaller effect sizes are present but do not reach statistical significance. This would be consistent with the finding that, for all significant parent--child interest relationships found among 16--18-year-olds, the direction of the relationship among 10--15-year-olds is the same --- but the coefficient is larger. Moreover, again confirming Hypothesis 2e, the largest effect sizes among 16--18-year-olds are for mother--daughter and father--son transmission. Substantively, when a topic is the one their mother discusses the most, 16--18-year-old daughters' interest in that topic increases on average by 1 point. Similarly, when a topic is the one their father discusses the most, 16--18-year-old sons' interest in that topic increases on average by 0.9 point.^[@fig-discussparentyo in @sec-appendix8 finds similar results when controls are removed, but the father--son link becomes significant among 10--15-year-olds.]

```{r}
#| label: fig-discussparentyoses
#| echo: false
#| fig-cap: "Interest in Topics by Gender, Age and Discussion with Parents, 2022 CPIS"
knitr::include_graphics("_graphs/DiscussParentYOSES.pdf")
```

## Discussion

Overall, the analysis generally confirms the main hypotheses put forward, with a few minor caveats. Broadly, it seems that social learning theory applies just as well when political interest transmission is evaluated across a range of political topics --- but not necessarily for each parent--child pair and topic measured individually. Moreover, whichever parent discusses a topic the most seems to matter for boys more than for girls.

Descriptive results show a clear trend: mothers are more likely to speak about education and health care, while fathers are much more likely to speak about the other three topics. These trends mirror previous literature on interest in these topics. Fathers tend to discuss the same topics with their sons as with their daughters, and the same goes for mothers, which supports previous findings that mothers and fathers discuss politics just as much with their children of both genders.

Moreover, fathers' discussion of various political topics is related to their sons' interest in these topics, while mothers' discussion of various political topics is related to their daughters' interest in these topics. Mothers also do not seem to influence sons' political interests, while fathers' influence over their daughters' political interests disappears when measured in a model alongside other socialization agents. Political interest seems to trickle down from parents to children in gendered ways.

Results by age groups may indicate political interest grows with time, but these results also seem to apply to girls more than to boys --- and the increase is not statistically significant. Future studies conducted among a larger number of children may be able to see more clearly time trends for the growth of parent--child political interest transmission potential.

The father--son link is particularly strong since when a boy's father discusses a topic more than their mother, he is likely to report more interest in that topic; whichever way it is measured, there seems to be a strong transmission link between fathers and sons, even though fathers discuss politics just as much with daughters and sons. Results highlight the transmission potential of parents for political interests to their children and they reinforce the idea that, in some respects, there are different socialization routes for girls and boys.