The Intergenerational Transmission of Party Preferences in Multiparty Contexts: Examining Parental Socialization Processes in the Netherlands

Research shows that parents have a strong influence on the party preferences of their children. Yet little is known about how such preferences are transmitted in multiparty systems with weak party identification and high electoral volatility. We propose a model of intergenerational transmission that includes both direct effects of parents’ party preferences on those of their children, as well as indirect effects through left– right and issue positions. We test this model with original survey data of Dutch adolescents (14– 20 years old) and their parents (N = 751 adolescent-parent pairs). We find two paths through which parents exert influence on the party preferences of their adolescent children. On the first path, parental party preferences function as a direct predictor of adolescent party preferences. On the second path, adolescent left– right and issue positions function as a mediator between parental left– right and issue positions and adolescent party preferences, with the effect of left– right positions being stronger than that of issue positions. The frequency with which adolescents discuss political topics with their parents moderates these effects.

Research on political socialization has repeatedly shown that people develop political values and preferences for political parties at a relatively young age, and that these values and preferences stabilize when they grow older (Alwin & Krosnick, 1991;Franklin, 2004;Jennings et al., 2009). It is therefore not surprising that scholars of electoral processes have studied how adolescents develop their political orientations. One source of information for adolescents learning about politics is their parents (e.g., Achen, 2002;Coffé & Voorpostel, 2010;Hooghe & Boonen, 2015;Jennings et al., 2009;Jennings & Niemi, 1968;Niemi & Jennings, 1991;Zuckerman et al., 2007), who transmit their party preferences to their children, leading to a high intergenerational similarity in preferences for political parties within families (Achen, 2002;Coffé & Voorpostel, 2010;Hooghe & Boonen, 2015;Jennings et al., 2009;Zuckerman et al., 2007).
Traditionally, most of the research on parental socialization of political orientations focused on the development of partisan orientations in countries with a two-party system such as the United States and the United Kingdom (Sapiro, 2004). Some recent studies focus on partisan orientations in multiparty systems like Switzerland and Belgium (Coffé & Voorpostel, 2010;Hooghe & Boonen, 2015). These studies demonstrate that even in multiparty systems, where party identification is generally weaker, there is a strong link between the party preferences of parents and their children. This is particularly the case in families where parents and their children discuss politics (Hooghe & Boonen, 2015). While these are important findings, it remains unclear how party preferences are transmitted in multiparty systems with high levels of electoral volatility such as the Netherlands.
We start from the general assumption that the political (and sociocultural) context partially determines which attitudes are transmitted in the family (see also Percheron & Jennings, 1981;Ventura, 2001;Westholm & Niemi, 1992). In two-party systems like the United States, where many citizens identify with one of the two parties, we expect that the party identification of parents can easily affect their children's party identification. Yet in a multiparty context where parents do not identify with one single party but switch easily between parties that are often ideologically similar, we expect the transmission process to operate (partially) indirectly through other orientations that have more heuristic value like left-right orientations (Ventura, 2001;Westholm & Niemi, 1992). To study parental transmission in a multiparty context, we therefore do not only focus on party preferences but also on the main determinants of party preferences, namely ideological and policy positions. We investigate direct and indirect effects of parents' political orientations on those of their children in the context of the highly fragmented Dutch multiparty system. Thus, our study speaks directly to calls for more research on political socialization processes in multiparty contexts (Sapiro, 2004).
We propose a model of parental socialization that integrates the relationships between the party preferences and the determinants of these preferences (left-right orientations, attitudes towards political issues) of both adolescents and their parents into a complex process of intergenerational transmission (see Figure 1). We posit that in addition to transmitting their party preferences to their children directly, parents also influence the party preferences of their children indirectly through the transmission of their left-right and issue positions. Moreover, we propose that the extent to which parents and their children discuss political issues influences the intergenerational transmission process. We employ structural equation modeling (SEM) to test this model, estimating the effects with original survey data collected in 2020 among Dutch adolescents (14-20 years old) and their parents (N = 751 adolescent-parent pairs).
We make three contributions to the literature about parental political socialization. First, our findings provide insights into the process through which parental party preferences are transmitted to their children, suggesting that in multiparty systems there is a direct relationship between the party preferences of parents and their children, but this relationship is also mediated (by leftright and issue positions) and moderated (by frequency of political discussions). Second, our findings show that the relationships between left-right orientations, attitudes towards political issues, and party preferences among children resemble those of their parents. Third, our findings demonstrate that, in the Dutch context, general left-right positions have a stronger effect than specific political issue positions in the socialization process. The strongest effects that we observe in the various models are from parents' left-right orientations on the left-right orientations of their children. This indicates that in the Dutch multiparty system parents first and foremost communicate general political values and orientations to their children. These conclusions support previous studies that focus on contextual differences in socialization processes (e.g., Ventura, 2001;Westholm & Niemi, 1992), and they are also relevant to research on broader changes in electoral trends, such as work on generational differences in electoral behavior (e.g., Andersen, 1976;Blais et al., 2004;Grasso, 2016;Grasso et al., 2019;Konzelmann et al., 2012;Tilley & Evans, 2014;van der Brug & Rekker, 2021;Walczak et al., 2012).

The Parental Socialization of Party Preferences
Research has shown that political orientations are formed early in life during the "impressionable" or "formative years" Flanagan & Sherrod, 1998;Hooghe & Wilkenfeld, 2008;Sears & Funk, 1999). This is the period ranging roughly from late adolescence to early adulthood (Abramowitz & Saunders, 1998;Andersen, 1976;Carmines & Stimson, 1981;Dalton, 2015;Hooghe, 2004). During the impressionable years, individuals are thought to be susceptible to influences from socialization agents that transmit political information, for example school, peers, family, media, and political events (Neundorf & Smeets, 2017). Through their interactions and experiences with these agents, adolescents learn about their political orientations and develop their own, which remain relatively stable afterwards (Alwin & Krosnick, 1991;Franklin, 2004;Jennings et al., 2009;Stoker & Jennings, 2008) and form the basis of further political learning. 1 Our article focuses particularly on the influence of one of these agents, namely parents and their effect on the party preferences of their adolescent children. Previous studies have shown the importance of examining intergenerational transmission of party preferences, that is, the effect of parental party preferences on those of their children (Achen, 2002;Coffé & Voorpostel, 2010;Hooghe & Boonen, 2015;Jennings et al., 2009;Zuckerman et al., 2007). Party identifications and party preferences that are transmitted through parental socialization have been found to be more stable over time compared to those that have not been transmitted by parents (Boonen, 2015;Jennings et al., 2009;Kroh & Selb, 2009). Much of this research stems from the United States, where party identification plays a central role in the process of political socialization. If parents identify strongly with a party, their children are likely to learn about these partisan orientations, so that a direct causal link between the party preferences of parents and their children is plausible.
However, in multiparty systems such as the Netherlands, few people identify with one single party (Thomassen & Rosema, 2009). Voters usually have consideration sets comprised of multiple parties with partly shared political positions, and these sets are relatively stable over time (Rekker & Rosema, 2019). From election to election, voters switch between parties in their consideration set, leading to high levels of volatility. van der Meer et al. (2015) estimate that more than half of the Dutch electorate switched their vote intentions at least once over the four years between the Dutch elections of 2006 and 2010. In this context, the direct transmission of party preferences from parents to their children is likely to be weaker than in two-party systems (Percheron & Jennings, 1981;Ventura, 2001;Westholm & Niemi, 1992). After all, it does not seem plausible that parents provide their children with strong and consistent cues about their party preferences when they themselves are torn between multiple parties.
Hence, we propose that the intergenerational transmission of party preferences may follow two routes. Via the first route, parental party preferences directly influence their children's party preferences without mediation. Via the second route, parental issue attitudes and especially their left-right orientations influence the attitudes and left-right positions of their children, and hence (indirectly) their party preferences. We base our latter expectation on recent empirical evidence that parents transmit more general political values, as expressed in their left-right orientations Ventura, 2001), as well as attitudes towards specific political issues (Kuhn et al., 2021;Meeusen, 2014) to their children. We would expect this latter path, especially the transmission of left-right identification, to be particularly important in the context of a multiparty system (Percheron & Jennings, 1981;Westholm & Niemi, 1992) in which people tend to switch easily between parties that are ideologically similar in terms of their left-right positions (e.g., van der Meer et al., 2015), indicating that citizens' consideration sets are strongly linked to their ideological positions. Moreover, these positions and consideration sets are more stable than their actual party choice. Therefore, we expect parents to be more likely to transmit their left-right orientations to their children than their actual party preferences. Notably, this increased intergenerational transmission of left-right identification in multiparty systems has also been explained with the heuristic advantage of left-right categorization (Ventura, 2001). Ventura proposed that both party identifications as well as left-right identifications (and also other categories such as party blocs) function as cues that individuals use to structure and simplify political landscapes. While party identifications are prominent in two-party systems, where nominal categorizations are sufficient, left-right identifications have higher heuristic value in fragmented and volatile multiparty systems. Parents, then, first and foremost use these advantageous cues when they are communicating with their children about politics and thereby transmit them. By transmitting these left-right orientations (as well as issue preferences) to their children, parents indirectly influence their children's party preferences. After all, these ideological and issue positions are strong predictors of party preferences. Hence, we propose a more integrated approach to intergenerational transmission that not only encompasses a single transmission path of political attitudes (e.g., either party preference or issue attitude transmission) but also examines the interplay between these different facets of political attitudes and their transmission. This approach will allow for a better understanding of transmission processes, especially in the context of multiparty systems.
Our integrated model is grounded in a policy voting model of party choice, which finds its origins in the work of Downs (1957). The basic assumption is that preferences of voters for political parties depend upon the ideological (left-right) distance between them, so that voters are expected to vote for the party with a left-right position that is closest to theirs. In studies among adults, the Downsian model of ideological (left-right) voting has often been applied to political issues as well (e.g., Ansolabehere et al., 2008;Carmines & Stimson, 1980;Lachat, 2011;Walczak et al., 2012). In models of party choice, left-right distance is usually the variable with the strongest predictive power in most Western European countries, including the Netherlands (e.g., Fuchs & Klingemann, 1990;Lachat, 2008;Oscarsson & Holmberg, 2016). Therefore, Dutch parents can be expected to give their children relatively strong and consistent cues about their left-right orientations and their attitudes towards different political issues.
Even though it has been well established that ideological and issue positions exert a strong effect on party choice among adults, there is little comprehensive research on attitudes towards political issues and left-right orientations as predictors of party preferences among adolescents, especially those below the voting age. This is surprising given the evidence that adolescence constitutes an important developmental phase for both issue attitudes (Rekker et al., 2015 and left-right orientations . The few studies explicitly focusing on the link between adolescents' issue attitudes and party preferences have shown that adolescents are able to match their issue attitudes to their party preferences to a certain degree (Bergh, 2013;Boonen et al., 2014;. Other studies that examined the association between partisan preferences and issue attitudes and included adolescents in their sample confirmed the relevance of adolescence for the development of this link . Previous studies also find that in multiparty systems, left-right orientations may be of higher relevance to the development of party preferences than party identification (Rico & Jennings, 2016;Ventura, 2001). Recently,  observed that in the Dutch multiparty system, compared to party identification, left-right orientations develop relatively early in life, are relatively stable over time, and show high degrees of intergenerational transmission. We build on these findings by extending the range of issue domains examined and including evaluations of a wide variety of parties as a measure of party preference (a conceptualization of partisanship that is more appropriate than party identification in the Dutch context; see Thomassen & Rosema, 2009). The integrated model that will be tested in this study is presented in Figure 1. We will test this model for left-right positions, as well as six political issue domains, which are also included in the Dutch National Elections Studies (van der Meer et al., 2017), since they have been shown to be important predictors of the vote in the Netherlands: immigration, government intervention in the economy, euthanasia, environmental protection, European integration, and income redistribution.
In addition to testing the hypothesized model, we examine whether the effects are stronger among adolescent-parent pairs that frequently discuss political topics with each other. Talking about politics is an important way in which parents can shape the political views of their children by communicating their own attitudes and preferences to them. Previous research has confirmed this notion, showing that talking about politics positively affects parent-child congruence in voting intentions (Hooghe & Boonen, 2015) and social and political attitudes (Meeusen, 2014;Meeusen & Boonen, 2020). Parental politicization facilitates the transmission process by making politics more salient and by increasing the likelihood of children picking up parental political cues (Dinas, 2014;Hatemi & Ojeda, 2020;Jennings et al., 2009;Jennings & Niemi, 1968;Wolak, 2009). Finding a facilitating effect of the frequency of political discussions would confirm that parent-child similarity in issue attitudes and party preferences within our model can at least partially be explained by an actual transmission process rather than status inheritance (Connell, 1972;Glass et al., 1986).

Context
Most research on political socialization has been conducted in two-party systems, where many parents identify strongly with one party. These can be seen as the most likely cases for the direct intergenerational transmission of party preferences. The data of the present study have been collected in the Netherlands, which has a system of proportional representation with a low electoral threshold and is one of the most electorally volatile countries in Europe (Mair, 2008). At the time of the data collection in 2020, 13 parties were represented in the Dutch Parliament. For the direct transmission of party preferences of parents to their children, this is probably a least likely case, because Dutch parents often switch their votes, and their children need to learn about many parties to understand their parents' (possibly ambiguous) party cues.

Data
We conducted an original online survey among Dutch adolescents (51.7% female, 47.8% male, 0.6% "other"; ranging from 14 to 20 years of age, M = 17.09 years, SD = 1.47 years) and their parents (62.3% female, 37.7% male; M = 49.17 years, SD = 7.56 years). The data were collected through three existing Dutch panels, the I&O Research Panel, the LISS Panel, and the CentERpanel (see Appendix S2 in the online supporting information for detailed information on the panels and sampling strategy). These panels are long-standing, scientific research panels. The LISS panel and the CentERpanel are household panels. Data collection started in June 2020 and was finalized in August 2020. 2 Prior to data collection, the survey was piloted among adolescents of the relevant age group, and with varying educational backgrounds, to exclude comprehension issues. In total, data on 751 adolescent-parent pairs was collected. 3

Measures
We measured party preferences by asking respondents to rate their propensity to vote ("On a scale from 0 to 10, rate how likely or unlikely it is that you will ever vote for these parties?") for the eight parties with the highest number of projected seats in the polls of May 2020 (Louwerse, 2020), namely the Christian Democratic Appeal (Christen-Democratisch Appèl; CDA), Democrats 66 (Democraten 66; D66), Forum for Democracy (Forum voor Democratie; FVD), GreenLeft (GroenLinks; GL), Labour Party (Partij van de Arbeid; PvdA), Party for Freedom (Partij voor de Vrijheid; PVV), Socialist Party (Socialistische Partij; SP), and People's Party for Freedom and Democracy (Volkspartij voor Vrijheid en Democratie; VVD). While there were 13 parties represented in the House of Representatives at the time of data collection, we needed to make a smaller selection of parties to limit survey length. Answer options ranged from "I will never vote for this party" (0) to "I will certainly vote for this party sometime" (10). An additional answer option was "I don't know this party." 4 Respondents were asked to indicate their own left-right orientation and estimate the leftright positions of the eight parties listed above 5 ("Where would you place yourself/party X on the following scale, where 0 means left, 5 means center, and 10 means right?"). Answer options ranged from "0 = left" to "10 = right." Moreover, respondents indicated their own position and estimated the positions of the eight parties listed above on six different statements (see Appendix S1 in the online supporting information). These statements address attitudes towards political issues on two core ideological dimensions that structure political preferences, namely the socioeconomic dimension and the sociocultural dimension (see e.g., Kriesi et al., 2008). More specifically, the six statements refer to attitudes towards immigration, government intervention in the economy, euthanasia, environmental protection, European integration, and redistribution of income. The statements were rated on 11-point scales and were formulated in such a way that a high score indicates a left-wing or liberal position on some issues and a right-wing or conservative position on others.
To measure the frequency of political discussions, adolescents were asked how often they talk about social and political issues with each of their parents ("How often do you usually talk about politics or societal issues with your [parent/caregiver]?"), while parents were asked this same question in relation to each of their children. Answer options ranged from "1 = Never" to "5 = Every day." Parents and adolescents agreed about their frequency of discussion almost all the time. The majority (87%) of adolescent and parent ratings are either the same (40%) or deviate from each other by one point (47%) on the 5-point response scale. A mean discussion frequency score was created by calculating the average of the score given by the adolescent in relation to the parent that was participating in the study with them, and the score given by the parent in relation to that adolescent. We tested whether using the discussion frequency rating given by the adolescent or the parent, instead of the average of the two, would change our findings. This was not the case.

Design
In the original data set, the responses of the adolescents and the responses of one of their parents were merged, so that they appear as a single observational unit. We refer to these units as "pairs" (adolescent-parent pairs). Because we have measured preferences (of parents as well as their children) for eight political parties and because we have information about the distance between these respondents and parties, we could estimate models for each of the parties separately. 4 The "voting propensity" measures were developed to study party choice in a multiparty system. Hence, they should not be seen as equivalent to measures of party identification (van der Eijk et al., 2006). 5 If respondents indicated not knowing a party early in the questionnaire, all later questions about parties did not include these unknown parties, and the responses were automatically coded as missing.
However, since our study does not focus on differences across party-specific models, we conduct analyses across all eight parties in a combined analysis. Thus, to analyze the relationships between adolescent and parental issue attitudes and voting propensities in a meaningful way, we restructured the dataset into a long format in two steps.
First, we created a format in which each case represents a pair*party combination. This means that for each of the adolescent-parent pairs, eight cases were created corresponding to ratings of the eight parties, leading to a total of 6,008 cases. Of these 6,008 cases, we deleted 931 cases (15.5%) due to item nonresponse. To estimate the relationship between issue attitudes and party preferences, we included the distances between a respondent's issue position and the position of the respective party on that same issue. Hence, if the unit refers to the combination between an adolescent-parent pair and Party A, the distances are between the position of the respondent on an issue and the position of Party A on that issue. To calculate these distance measures, we used mean party-position ratings of parents who gave correct answers to all of three political-knowledge questions. 6 We computed the distances in relation to the same "objectified" party position for parents and children, so that the correlations are only the result of similarities in positions of parents and children, not affected by the variations in perceptions of party positions. Moreover, the perceptions of party positions may be influenced by party preferences as well. Even though reciprocal relations might exist between party preferences and issue distances, this limits the potential bias. Parents' responses were used as they can be expected to be on average more familiar with the party system. To estimate the model for issue attitudes, we also created a format in which each case represents an adolescent-parent pair*party*issue combination. This means that for each of the adolescent-parent pair*party cases, six cases were created corresponding to the six issues that were rated by respondents, leading to a total of 36,048 cases. In order not to artificially inflate the sample size, we weighted the units in the "stacked" data matrix, so that the significance tests are based on the original sample size. 7 Once the data was restructured, we tested our hypothesized model through SEM with maximum-likelihood estimation. We used the first format (adolescent-parent pair*party) to test the model separately for all six issue dimensions and for left-right orientations, and the second format (pair*party*issue dimension) to test a "metamodel" including attitudes towards political issues more generally. We evaluated the goodness of fit of our models using the root mean square error of approximation (RMSEA) and the comparative fit index (CFI) as recommended by Kline (2015). We define good model fit by the following criteria: RMSEA ≤.05, with an upper 90% CI bound of ≤.10; CFI ≥.95. We also examined the modification indices for each model. The path model was estimated using IBM SPSS AMOS 25.
During the second stage of our analyses, we examined the moderating effect of discussion frequency. To do so, we used a median split to divide the adolescent-parent pairs into one group that reported talking to each other about political issues with low frequency and one that reported doing so with high frequency. 6 Previous research has shown that citizens' mean ratings of party positions are good approximations of actual party positions (van der Brug & van der Eijk, 1999; Simas & Evans, 2011). Expert ratings were not used as they were not available for these specific items. 7 If each pair appears eight times, in combination with eight parties, the cases get a weight of 1/8. If each pair appears 48 times, in combination with six issues and eight parties, the weight is 1/48.

Model Testing and Modification
Tables 1 and 2 provide descriptive information about the core variables in our models, as well as the correlations between these variables. These correlations are the input of the SEM analyses, which show significant effects for all hypothesized paths in the models for the intergenerational transmission of left-right orientations (see Figure 2, Model A) and attitudes towards political issues (see Figure 3, Model A). This also holds true for the six separate models for each issue dimension (see Table 3, Model A; for reasons of space, the path diagrams of these models are presented in Appendix S3 in the online supporting information), with one exception in the model including attitudes towards government intervention (insignificant effect of adolescent attitudes on adolescent party preferences). This means that in both the general models and all issue dimension-specific models, parental left-right and issue positions have a significant effect on adolescent voting propensities. This effect is mediated by adolescent left-right and issue positions and also by parental voting propensities. Overall, the models show an acceptable fit based on the RMSEA and CFI indices. Still, the RMSEA index is slightly higher than our set cut-off for the left-right and issue position models and considerably higher than our set cut-off in two issue domain-specific models (see Table 3, immigration and European integration, Model A). The modification indices for these models suggest that adding a path from adolescent left-right and issue positions to parental party preferences would improve their fit. The modified models can also be seen in Model B in Figures 2 and 3 and Table 3. The modified left-right and issue-position models again show significant effects for all hypothesized paths, as well as a significant effect of adolescent left-right orientations and issue attitudes to parental voting propensities in the general model and the models for attitudes towards immigration and European integration. The goodness of fit is improved, and the models fit well in all instances. The path effects show a considerable influence of parents on their children. Parental issue attitudes as well as party preferences have a strong positive effect on the left-right orientations and issue attitudes and party preferences of the adolescents. In Appendix S6 in the online supporting information, we report predicted values, which provide another indication of the sizes of the effects. Notably, the intergenerational transmission effects are stronger for left-right orientations than for issue attitudes. Moreover, parents' party preferences and left-right and issue positions appear to have a stronger effect on adolescents' party preferences than adolescents' own left-right and issue positions. Nevertheless, even in this relatively young sample, adolescents' left-right orientations and issue attitudes considerably affect their own party preferences,   and this association is significant across all issue dimensions except government intervention. Relative to the other issue dimensions, the associations between adolescents' attitudes towards immigration and European integration and their party preferences are especially strong. This is also true for the association between adolescents' attitudes towards immigration and European integration and the party preferences of their parents. While this effect appears overall considerably smaller than effects of parental attitudes on adolescent voting propensities, it remains robust in the general model and in the issue-domain specific models for attitudes towards immigration and European integration.
In sum, the path analyses for all models provide clear support for the hypothesized effects of parental left-right orientations and issue attitudes on adolescent voting propensities, mediated by adolescents' left-right orientations and issue attitudes and parental voting propensities. These effects can be classified as moderate to strong, explaining a considerable amount of variance in adolescent voting propensities (33% and 29% in the general models for left-right orientations and issue attitudes respectively). Remarkably, we also find a direct effect of adolescent left-right and issue positions on parental voting propensities.

Discussion Frequency as Moderator
The path analyses of the modified general models conducted separately for adolescentparent pairs with low versus high frequency of discussing political issues can be seen in Figures 4 and 5. As expected, the group that reported discussing social and political issues relatively more frequently shows stronger positive effects of parental on adolescent left-right and issue positions and voting propensities. This suggests that increased discussion frequency strengthens the intergenerational transmission of parents' left-right and issue positions and voting propensities to their adolescent children. The models continue to show good global fit after the inclusion of discussion frequency as a moderator.

Other Moderators
In Appendix S4 in the online supporting information, we report detailed results of other moderating factors. First, we tested whether differences in adolescent or parental political knowledge moderate the observed effects. Surprisingly, however, no moderating effects of adolescent or parental political knowledge on intergenerational transmission paths were found. Yet the association between left-right and issue positions and voting propensities is higher among politically knowledgeable adolescents. The same is true for parents high in political knowledge. This highlights the importance of political knowledge in building a coherent political belief system, both in adolescence and adulthood.
Second, we examined whether the observed path effects were moderated by adolescent or parental political engagement. The results indicate that high parental political engagement positively influences the transmission effects from parents to children (issue attitudes and voting propensities), whereas high adolescent political engagement positively affects the paths from adolescents' left-right orientations and issue attitudes to parental voting propensities.
Third, we compared the strength of parental transmission from mothers and fathers to their children. The transmission effects are stronger in those cases in which a mother filled in the survey than in cases in which the respondent was a father. Existing research shows that such differences can be caused by greater feelings of relative closeness between children and their mother, as well as more frequent interactions between mothers and their children (see Jennings, 1983;Jennings & Langton, 1969;Zuckerman et al., 2007). One of these explanations does not apply in our case: The fathers in our sample talk on average more about politics with their children than the mothers. Our data do not enable us to test the other suggested causal mechanism, namely that mothers have a stronger bond with their children than fathers. Thus, while our findings are the same as in previous studies-a stronger transmission effect from mothers than from fathers-we cannot draw conclusions about the mechanisms causing these differences.

Robustness
In Appendix S5 in the online supporting information, we estimated our models using respondents' own perceptions of party positions to compute the measures representing the distance between parties and respondents on the left-right and issue dimensions. Doing so weakens the estimated transmission effects and inflates the estimated effects of left-right and issue positions on party preferences. Yet the overall conclusions do not change.

Discussion
We put forward and tested a model that integrates the link between adolescents' left-right orientations, issue attitudes, and party preferences into a more complex process of intergenerational transmission. We find two paths through which parents influence the party preferences of their adolescent children. On one of these paths, adolescent left-right and issue positions function as a mediator between parental left-right and issue positions and adolescent party preferences. This means that parents partially transmit their own left-right and issue positions to their adolescent children, and these transmitted positions in turn influence their children's party preferences. On the second path, parental party preferences function as a mediator between parental left-right and issue positions and adolescent party preferences, meaning that parents' left-right and issue positions affect their own party preferences, and these parental party preferences then are partially transmitted to their adolescent children. Thus, our first conclusion is that, in complex multiparty contexts like the Netherlands, parental socialization operates to a large extent through the transmission of policy preferences and particularly left-right orientations (see also Percheron & Jennings, 1981;Ventura, 2001;Westholm & Niemi, 1992).
A second important conclusion is that left-right and issue positions of young people (most of whom are not yet eligible to vote) are already sufficiently developed to significantly contribute to their party preferences. While the strength of the relationship between left-right orientations, issue attitudes, and voting propensities seems overall weaker for adolescents than for their parents, these effects are robust across all issue domains. That the determinants of party preferences have a higher impact for older compared to younger individuals makes sense and is in line with previous findings on political learning and life cycle effects (van der Brug & Rekker, 2021;Stoker & Jennings, 2008). After all, more mature voters can be expected to have a clearer view on their own and different parties' positions on the left-right spectrum and political issues than the youngest ones. Our study constitutes a stringent test of the link between party preferences and their determinants, seeing as the adolescents in our sample are politically socialized in a multiparty context. Consequently, they need to not only have sufficiently developed left-right orientations and issue attitudes but also to learn about the positions of a high number of parties on these predictors for there to be a detectable link between the two. This implies considerable sophistication of adolescents' political beliefs, confirming prior studies that have demonstrated adolescents' ability to connect their own attitudes to their vote choices (Bergh, 2013;Boonen et al., 2014;. These findings are also relevant in view of studies showing that adolescents rely increasingly on their own political attitudes during their political socialization. This has been interpreted as a sign of diminishing parental influence (Sears & Valentino, 1997). A stronger connection between different facets of political orientations could indeed be regarded as a sign of a more coherent and sophisticated belief system and therefore as signaling maturation. However, we show that the role of parents remains profound, particularly because parents do not only influence their children's party preferences but also exert considerable indirect influence on their children's party preferences by influencing their left-right orientations and issue attitudes. The long-term transmission effects of parents on adolescent left-right orientations and issue attitudes may be better predictors of the future electoral behavior of adolescents, simply because these orientations and attitudes may show higher stability over time (Hooghe & Wilkenfeld, 2008). Moreover, in fragmented multiparty systems such as the Netherlands, party landscapes change frequently, hampering the transmission of party preferences. Thus, while the parties that a parent liked (or disliked) during the formative period of their child may cease to exist at some point, their left-right orientations and attitudes towards political issues will continue to influence their offspring's evaluations of newly emerging parties.
The Dutch multiparty context is somewhat unique in its fragmented nature and absence of large parties. However, most European countries have multiparty systems, and fragmentation seems to be increasing in other countries as well. A large number of parties and low degrees of party identification generate large consideration sets of voters, and high electoral volatility leaves a lot of room for ambiguity in cue recognition. Hence, the effects of parental party preferences on those of their children can be expected to be weaker in multiparty systems than in two-party systems. Nevertheless, we find strong associations between the political preferences of Dutch parents and their children. These associations are stronger for general left-right orientations than for attitudes towards specific political issues. This finding concurs with previous research suggesting that left-right orientations are the strongest predictors of electoral preferences in most Western European countries, including the Netherlands (e.g., van der Brug, 2010). Given its importance in determining parental party preferences, it seems plausible that parental political socialization in multiparty systems occurs largely through the transmission of general political values. These results thus support the more general theoretical idea behind this article, that the political (and sociocultural) context partially determines which attitudes are transmitted in the family (see also Percheron & Jennings, 1981;Ventura, 2001;Westholm & Niemi, 1992).
Furthermore, among adolescents and parents who frequently discuss politics with each other, we find higher similarities in left-right orientations, issue attitudes, and voting propensities. This is likely explained by a higher chance of adolescents learning about the political views of their parents before adopting them. We considered the possibility that the moderating effect of discussion frequency would be due to higher political knowledge. However, neither parental nor adolescent political knowledge has a moderating effect on the transmission paths. Thus, it appears that the moderating effect of discussion frequency ratings is indeed a consequence of the amount of communication about politics between adolescents and their parents. Notably, discussing political topics with each other does not appear to be the only mode of communication that affects the intergenerational transmission of political preferences. We find that politically engaged parents and adolescents show a higher similarity to each other in their left-right and issue positions and their party preferences, probably because they are likelier to communicate their political preferences to the other nonverbally (e.g., by campaigning or wearing clothes with political messages).
Unexpectedly, we also find that the left-right and issue positions of adolescent children influence the party preferences of their parents. This finding is in line with scholarship emphasizing the importance of reciprocal or trickle-up influence (McDevitt & Chaffee, 2002;Zuckerman et al., 2007) and children's agency in the socialization process (Hatemi & Ojeda, 2020;Ojeda & Hatemi, 2015). In our transmission model, the effect that children have on their parents is overall much smaller than vice versa. This could be explained by the authoritative role parents hold in their families, both in terms of social hierarchy as well as political experience. The relative effect of transmission paths may change, however, in families where children are more politically knowledgeable and engaged than their parents or possess more political self-efficacy (Wong & Tseng, 2008). In these constellations, a trickle-up dynamic where children teach their parents about politics may be more prominent.
In this regard, our findings further highlight research into bidirectional intergenerational transmissions as an interesting avenue for future research. An obvious limitation of the present study is the cross-sectional nature of the data utilized. The associations we find, paired with the enhancing effect of communication between parents and adolescents, as well as good model fit, point towards causal processes. Yet to further disentangle the reciprocal effects between political orientations of parents and their children, future studies would require longitudinal data (York, 2019).