Political Socialization and Voting

Introduction

Brady, Schlozman, and Verba (2015) have recently drawn attention to the lack of study of political reproduction. Just as social reproduction refers to the processes that transmit social and economic inequality from one generation to the next, political reproduction denotes the processes underpinning the intergenerational transmission of political inequality. Brady et al. argue that political reproduction and social reproduction are inextricably linked: parental socioeconomic disadvantage translates into political disadvantage. When it comes to transmitting political inequality, the key aspect of socioeconomic status (SES) is seen to be parental education. According to status transmission theory, education serves as “the engine for the transmission of political activity from generation to generation” (Verba, Schlozman, and Burns 2005, 98). Well-educated parents are more likely to provide a politically stimulating home environment and, more importantly, they are more likely to have children who are well educated.

Status transmission theory poses a challenge to the social learning perspective that has dominated studies of political socialization. Social learning theory highlights the role of observational learning and the modeling of behavior on the parental example (Bandura 1977; see, for example, Jennings, Stoker, and Bowers 2009). The two theories are not necessarily incompatible as observational learning could be one path through which the parents’ educational attainment influences the propensity of their adult children to be politically active. However, according to proponents of status transmission theory, this path is secondary as the critical mechanism behind the reproduction of political inequality is the parent–child link in educational attainment (Verba, Schlozman, and Burns 2005).

Status transmission theory is an important contribution to our understanding of political inequality in the United States, but there is a question mark over the generalizability of the theory to other settings. The U.S. context is not typical when it comes to the intergenerational transmission of educational attainment. Among thirteen high-income Western countries, only Italy (.54) has a higher intergenerational schooling correlation (Hertz et al. 2007). This puts the United States (.46) on a par with Ireland and Switzerland, but well ahead of countries like Denmark (.30), Great Britain (.31), New Zealand (.33), Norway (.35), and the Netherlands (.36). In fact, as far as the correspondence between the education of parents and their children is concerned, the United States is much closer to countries like Pakistan (.46), Sri Lanka (.48), and Egypt (.50) than to most of its high-income counterparts.

To assess the generalizability of status transmission theory, we investigate the relative importance of status transmission and social learning in explaining the intergenerational transmission of unequal turnout in Finland. There are a number of reasons why Finland is a good case for this purpose. First, Finland (.33) ranks thirty-fifth among forty-two countries in terms of the average parent–child schooling correlation, whereas the United States ranks fifteenth (Hertz et al. 2007). ¹ There is, nonetheless, an important element of educational persistence:

. . . in Finland too, the universal logic that the children of parents with a high level of education will have a long educational career and that the children of parents with a low level of education will follow a shorter and more practically oriented path applies. (Kivinen and Rinne 1996, 303)

At the same time, Finland is one of the countries, along with the United States, where education has a particularly strong relationship with turnout, especially among young people (Gallego 2015). In other words, Finland is not a least-likely case for status transmission theory to hold.

A final advantage of studying Finland is that it enables us to overcome one of the key limitations of much of the existing literature, namely, the reliance on recall data (Brady, Schlozman, and Verba 2015). Our analyses take advantage of a unique data set that links individual-level voting records from the 1999 Finnish parliamentary elections with data on socioeconomic background characteristics compiled by Statistics Finland for a 10 percent sample of eighteen- to thirty-year-old voters. Crucially, personal identification numbers enable information on the turnout and education of both parents to be matched to their offspring with a high degree of accuracy. Using actual voting records means that our data avoid the biases associated with self-reported turnout, such as misreporting due to faulty recall or overreporting due to social desirability (Karp and Brockington 2005; Swaddle and Heath 1989). Misreporting is a particular problem when studies of intergenerational transmission have to rely on respondents’ recollections of their parents’ political activities. At the same time, using official data minimizes the problem of inaccurate reporting of parental education.

Status Transmission Theory and Unequal Turnout

Status transmission theory highlights the role of parents’ SES, and more particularly their educational attainment in perpetuating unequal participation in politics (Schlozman, Verba, and Brady 2012; Verba, Burns, and Schlozman 2003; Verba, Schlozman, and Burns 2005). Verba, Burns, and Schlozman (2003, 58) argued that

. . . in politics, just as in economics, the driving force behind intergenerational transmission of inequality is parents’ education. Parental education has a potent impact on political activity because well-educated parents are more likely both to provide a politically rich environment and to have children who become well-educated, affluent adults.

There are thus two mechanisms through which well-educated parents influence their children to be politically active adults. The first mechanism is more direct: children whose parents are well educated are more likely to be exposed to politics in the home environment. Verba and colleagues (Verba, Burns, and Schlozman; 2003; Verba, Schlozman, and Burns 2005) found that adults who recall their parents discussing politics and engaging in political activities are more likely to be politically active themselves. The second and more important mechanism is indirect: high-status parents tend to pass their socioeconomic advantage on to their children and this socioeconomic advantage translates into political advantage. From this perspective, the critical aspect of the parents’ SES is their level of education. This influences the level of education attained by their children. Education paves the way for well-paying careers in the sorts of occupations that encourage the development of civic skills, as well as membership in politically active social networks. Education also fosters interest in politics, political knowledge, and a sense of political efficacy. As a result, well-educated parents are more likely to have children who are politically active as adults. ² Parental education, not parental example, is the driving force in status transmission theory: “the learning process emphasized in the literature on political socialization is of secondary importance” (Schlozman, Verba, and Brady 2012, 186; see also Verba, Burns, and Schlozman 2003; Verba, Schlozman, and Burns 2005).

Social Learning Theory and Unequal Turnout

Social learning theory, on the contrary, emphasizes the importance of observational learning:

. . . most human behavior is learned observationally through modeling: from observing others one forms an idea of how new behaviors are performed, and on later occasions this coded information serves as a guide for action (Bandura 1977, 22)

In this conception, parents act as models whose behavior children encode and later come to imitate. This is very much a matter of learning by example. Modeling can be based on live actions. For example, a child may be influenced by seeing a parent going to vote or by accompanying the parent to the polling place. Modeling can also be based on verbal instructions. For example, parents may actively teach their children that citizens have a civic duty to vote or they may emphasize the expressive value of voting. Others may teach lessons about the importance—or not—of voting based on a weighing of costs and benefits. These lessons can influence turnout. A Finnish study found that people who had received advice from their parents about the importance of voting were significantly more likely to vote (Wass 2007a). When the child becomes eligible to vote, verbal instructions from parents can lower the start-up costs of voting (Plutzer 2002). Parents can help the first-time eligible voter get registered to vote and explain what happens in the voting booth, and they may accompany the young voter to the polling station.

Observational learning is most effective when the behavior of models is consistent (Bandura 1977). This suggests that intergenerational transmission will be strongest when there are consistent parental cues. If so, young adults will be most likely to vote when both of their parents are voters. They will be less likely to vote when one parent votes but the other parent does not. In this situation, the parental cues will be inconsistent and the influence of the nonvoting parent is likely to diminish the effect of having a parent who votes.

When only one parents votes, which parent is the voter is likely to matter. Social learning theory emphasizes that the effectiveness of observational learning depends on characteristics of the role model (Bandura 1977). When social learning theory was applied to the intergenerational transmission of political trust, the theory proved to be more successful in explaining the mother–child link than the father–child link (Campbell 1979). This is consistent with other studies of political socialization, which have typically found that the mother’s influence outweighs the father’s (Acock and Bengtson 1978; Bhatti and Hansen 2012a; Oberle and Valdovinos 2011; Shulman and DeAndrea 2014; but see Abendschön 2013). Acock and Bengtson (1978, 528) went so far as to argue that “if one parent is to be used to represent family influence in socialization, in most cases it should be the mother.” Mothers’ influence has traditionally been explained in terms of their role as primary caregivers (Oberle and Valdovinos 2011). From the perspective of social learning theory, though, sex differences in the style of verbal modeling may also matter. Communications scholars have attributed the difference in fathers’ and mothers’ political influence to the way that they typically speak with their children. Shulman and DeAndrea (2014) concluded that mothers are more likely to influence their children’s political affiliations because they are more likely than fathers to encourage open discussion rather than trying to impose their own views. These studies have not explored the relative influence of fathers and mothers on the turnout of their adult children, but it seems likely that similar processes will come into play.

It is possible, though, that the parents’ influence depends on the child’s sex. According to social learning theory, one of the key factors governing the effectiveness of the modeling process is observer-model similarity (Bandura 1969). Children are more likely to model their behavior on the behavior of those who are perceived to be similar to them. In the context of the family, this means that children will tend to model the behavior of a parent of the same sex (Bussey and Bandura 1999). Hence, we would expect mothers to have a stronger impact than fathers on the turnout of their daughters whereas fathers have more influence than mothers on their sons. Studies of political socialization have consistently shown that the mother–daughter link holds (see Atkeson and Rapoport 2003; Gidengil, O’Neill, and Young 2010; Owen and Dennis 1988; Rapoport 1985), whereas the father–son association has proved less consistent. Rapoport (1985) found that fathers had a somewhat stronger influence than mothers on their sons’ opinion expression, though the difference between the fathers’ and the mothers’ influence was much smaller for sons than for daughters. However, other studies have reported that both parents appeared to influence their sons to more or less the same extent (Owen and Dennis 1988) or that neither parent influenced their sons’ willingness to express political opinions (Atkeson and Rapoport 2003).

A critical question about the parental example is whether parental influence persists once the adult child leaves home. On “leaving the nest,” young adults may come under the influence of peers who are less likely than their parents to be voters (Bhatti and Hansen 2012a). Even if they live with their parents into their mid- or late-twenties, their peer networks are likely to expand as they move on to college or join the workforce. As a result, parental influence may diminish. However, social learning theory suggests that the influence of peers will not necessarily outweigh that of parents: “The anticipation of self-reproach for conduct that violates one’s standards provides a source of motivation to keep behavior in line with standards in the face of opposing inducements” (Bandura 1977, 154). This notion of self-regulation is one of the central tenets of social learning theory. In the context of turnout, the standards of behavior that people set for themselves may derive from a sense of civic duty and the self-reproach may take the form of guilt for having failed to fulfill their duty to vote. Indeed, many voters report that they would feel guilty if they did not vote (Blais 2000). To the extent that parental modeling includes learning norms of civic duty, we could expect that self-regulation will ensure that many adult children of voting parents continue voting once they have left home. A similar argument applies to age.

Assessing Status Transmission Theory and Social Learning Theory

Status transmission theory implies that a substantial portion of the impact of parents’ education on the turnout of their adult children will be explained by parental turnout and by the child’s educational attainment, with the latter being the more important mediator (Hypothesis 1 [H1]). From this perspective, social learning theory simply offers a way of fleshing out the first mechanism specified by status transmission theory. To demonstrate that social learning theory has independent explanatory power, we need to show that the relationship between parental turnout and the turnout of young adults holds, even controlling for parental education (Hypothesis 2 [H2]). Given the importance of consistent cues and observer-model similarity, we should also find that the impact of parental voting is weaker when only one parent votes (Hypothesis 3 [H3]), and that when only one parent votes, the influence of the same-sex parent on the adult child’s turnout is stronger than that of the opposite-sex parent (Hypothesis [H4]). Furthermore, the importance accorded to self-regulation suggests that the impact of having parents who both vote should persist even when the adult child no longer resides with the parents and will keep them voting well into their twenties (Hypothesis [H5]).

Data and Research Design

Focusing on the relationship between the turnout of parents and their adult children has the advantage of enabling us to analyze hard data on turnout rather than having to rely, like Verba and his colleagues, on adults’ recall of their parents’ political activity and the frequency of political discussion within the home. Our analyses are based on a unique data set that links the voting records of young adults and their parents to socioeconomic characteristics provided by Statistics Finland. The turnout data are based on a 10 percent sample of eighteen- to thirty-year-olds randomly drawn by Statistics Finland from individual-level register records for the 1999 parliamentary election. ³ The data cover the mainland Finnish electorate. The voting information is collected from electoral wards and linked with registration data using personal identification numbers. The linkage rate is more than 99.5 percent. ⁴ Voting information on parents and their children is also combined through personal identification numbers. Parent–child linking was not possible in cases where the parent was deceased, unknown, a foreign citizen, or living in one of the handful of municipalities where voting records were not available. For these reasons, the father’s voting information could not be linked for 11,145 cases and the mother’s voting information could not be linked to their child for 5,117 cases, leaving an effective sample size of 67,465. ⁵ In 2,977 cases, more than one child of the same parents is included in the data set. To take this intrafamily correlation into account, we use robust standard errors clustered by family identification number.

We start with status transmission theory. Using the method developed by Karlson, Holm, and Breen (2012; see also Breen, Karlson, and Holm 2013), we compare the extent to which parental voting and the adult child’s own educational attainment mediate the relationship between parental education and the propensity to vote. This approach extends the decomposition properties of linear path models to logit and probit models, making it possible to estimate total, direct and indirect effects when the outcome variable is binary. ⁶ Like any method for identifying causal mediation effects, it is only valid under the assumption of sequential ignorability (Imai et al. 2011). To give mediation effects a causal interpretation, we must assume, first, that there is no confounding between the treatment variable and the mediating and outcome variables and, second, that there is no confounding between the mediating variables and the outcome variable. Sequential ignorability is necessarily a strong assumption that we approximate by controlling for three potential confounders: age, sex, and mother tongue. ⁷ Sequential ignorability also requires that neither mediator confound the relationship between the other mediator and the outcome. This is a reasonable assumption in our case as parental turnout is not going to influence the child’s educational attainment and vice versa. Both parental education and the adult child’s education are coded as continuous variables (cf. Highton and Wolfinger 2001). These consist of five categories, ranging from junior high school to graduate degree, whereas parental voting has three categories (neither parent voted, one parent voted, both parents voted).

The next set of models tests social learning theory and its implications. We estimate the impact of parental voting on the turnout of young adults with controls added first for parental education and then for parental income and finally for the education and income of the adult child (as well as the aforementioned controls). If the relationship between parental voting and the turnout of their adult children is attributable to parental education, controlling for parental education should significantly reduce or eliminate the original relationship. To test hypotheses about the relative influence of the mother versus the father, parental voting is represented by three dummy variables: both parents voted, only the father voted and only the mother voted, with neither parent voted serving as the reference category. Interactions between parental voting and the young adult’s sex, age, and living arrangement are used to test the remaining hypotheses.

For ease of interpretation, the regression results are presented in the form of marginal effects, estimated using the observed-value approach (Hanmer and Kalkan 2013). ⁸ Comparing marginal effects is important as even small effects are likely to achieve conventional levels of statistical significance with a sample this size. As a robustness check, the analyses were repeated on random split halves. The observed patterns were reproduced. ⁹ As an additional robustness check, we used matching prior to estimating the logistic regression models. Matching offers a potentially more powerful way of achieving control. Rather than assuming that potentially confounding variables affect turnout in a simple additive manner, matching mimics the logic of experimentation by explicitly comparing treatment (e.g., both parents voted) and control (e.g., neither parent voted) cases with the same values on all of these variables. As the cases have the same values, those variables cannot explain any difference in turnout. Depending on the comparison, there can be many unmatched cases. As a result, the data set of matches varies across the comparisons, which raises issues for drawing inferences. For example, the group yielding matches for neither parent voted is likely to be different depending on whether the cases are being matched with cases where both parents voted or with cases where only the mother voted. Accordingly, we have opted to present the results without matching, which are generally very similar to those obtained using matching. ¹⁰ The results with matching can be viewed in the online appendix (http://prq.sagepub.com/supplemental/).

Findings

Parental education clearly has an impact on the turnout of their adult children (see Table 1). Equally clearly, that impact is partly mediated by both parental voting and the young adult’s own educational attainment. Together, these two variables mediate 44.4 percent of the effect of parental education. Thus, as predicted by status transmission theory, a substantial portion of the impact of parents’ education on the turnout of their adult children can be explained by parental turnout and by the child’s own educational attainment (H1). However, contrary to our first hypothesis, it is the child’s educational attainment that is of secondary importance. Parental voting is the more important mediator. Indeed, parental voting mediates more than twice as much of the effect of parental education (30.1%) as does the child’s own educational attainment (13.3%). This finding indicates that learning may, in fact, be of primary importance in a context where the parent–child schooling correlation is not as strong as it is in the United States.

OPEN IN VIEWER

Table 1.

Decomposing the Total Effect of Parental Education on Voting into the Direct Effect and the Indirect Effects via Child’s Education and Parental Voting.

Total effect	0.322 (0.007)***
Direct effect	0.179 (0.007)***
Indirect effects
Via child’s education	0.046 (0.002)***
Via parental voting	0.097 (0.002)***
Mediation percentages
Total	44.4%
Via child’s education	13.3%
Via parental voting	30.1%

The decomposition was performed on 67,465 cases using the KHB method (Karlson, Holm, and Breen 2012). The model was adjusted for age, sex, and mother tongue. The standard errors are shown in parentheses and were clustered by the family identification number.

***

p < .001.

This conclusion is reinforced when we look at the impact of parental voting (Table 2). The estimated turnout of an adult child whose parents both vote is fully 39.3 points higher than that of an adult child with nonvoting parents. Strikingly, the magnitude of parental influence is only slightly diminished when controls are added for parental education and remains strong even with additional controls for parental income and the adult child’s income and educational attainment. In other words, the intergenerational transmission of SES can explain only a very small part of the relationship between parental voting and the turnout of their adult children. Even controlling for both parental SES and the adult child’s SES, the estimated turnout of young adults is 34.4 points higher when both parents voted than when neither parent did. Accordingly, we can conclude that our second hypothesis is resoundingly supported: the parent–child link in voting is clearly not spurious. Parental voting has an independent effect on the turnout of young adults, as predicted by social learning theory (H2).

OPEN IN VIEWER

Table 2.

The Estimated Impact of Parental Voting on Their Adult Child’s Turnout.

	Parental voting	Controlling for parental education	Controlling for parental and children’s education and income
Only father voted	13.8 (0.01)***	13.0 (0.01)***	12.6 (0.01)***
Only mother voted	21.8 (0.01)***	20.7 (0.01)***	20.4 (0.01)***
Both parents voted	39.3 (0.01)***	36.5 (0.01)***	34.4 (0.01)***
No. of cases	67,465	67,465	67,465

Column entries indicate the estimated change in the child’s probability of voting compared with the reference group of “neither parent voted.” Robust standard errors, clustered by family identification number, are shown in parentheses. The estimations are based on logistic regression, with controls for sex, age, mother tongue, marital status, housing tenure, and residential mobility.

***

p < .001.

As predicted by social learning theory (H3), it clearly makes a difference whether one parent or both parents vote (Table 2). Adult children are much less likely to vote when only one parent votes, suggesting that the example of the nonvoting parent partially offsets the influence of the parent who votes. Nonetheless, having at least one parent who votes does boost turnout. Our expectations regarding the relative influence of fathers and mothers are also supported. Even allowing for possible differences in their income and education, turnout is 20.5 points higher if the mother is the only parent who votes but only 12.6 points higher if it is the father. The difference in these marginal effects is statistically significant, indicating that the mother’s influence outweighs the father’s. This pattern also holds when matching is used prior to estimating the models (see the online appendix at http://prq.sagepub.com/supplemental/).

When interactions between parental voting and the adult child’s sex are added, however, it becomes clear that there are sex-differentiated effects. Based on social learning theory, we had anticipated that mothers would have a stronger impact than fathers on the turnout of their adult daughters whereas the reverse would hold for fathers and sons (H4). This hypothesis was based on the assumption that children are more likely to model their behavior on the behavior of those who are perceived to be similar to them. Observer-model similarity clearly matters in the case of mothers and daughters: when only one parent votes, having a mother who votes is more consequential for the daughter’s turnout than having a father who votes (Table 3). However, the converse does not hold: if anything, the fact that the father votes matters less for his adult son’s turnout than having a mother who votes, though this difference in marginal effects is of borderline statistical significance. It should be noted, though, that fathers do have significantly more influence on the turnout of their adult sons than on the turnout of their adult daughters whereas the mother’s influence is not significantly different for sons and daughters.

OPEN IN VIEWER

Table 3.

The Estimated Impact of Parental Voting on Their Adult Child’s Turnout According to the Sex of the Child.

Male
Only father voted	15.0 (0.01)***
Only mother voted	18.9 (0.01)***
Both parents voted	36.0 (0.01)***
Female
Only father voted	10.2 (0.01)***
Only mother voted	22.1 (0.01)***
Both parents voted	33.0 (0.01)***

Column entries indicate the estimated change in the child’s probability of voting compared with the reference group of “neither parent voted,” estimated on the basis of logistic regression models, with controls for age, mother tongue, marital status, education, income quartile, housing tenure, residential mobility, parental education, father’s income quartile, and mother’s income quartile, along with sex and interactions between sex and parental voting. Robust standard errors, clustered by family identification number, are shown in parentheses.

***

p < .001.

The importance accorded to self-regulation in social learning theory led us to expect that the impact of having parents who both vote would persist even when the adult child no longer resides with the parents and regardless of the age of the young adult (H5). Two-thirds of the eighteen- to thirty-year-olds in our sample had left home. Having parents who both vote is associated with significantly higher turnout on the part of young adults, regardless of whether or not they live at home (Table 4). Estimated turnout is 65.7 percent when both parents vote and the adult child lives with both parents, compared with 60.4 percent if the adult child lives with neither parent. It may be that so many continue to vote because they have paid the start-up costs of voting (Plutzer 2002), but given the importance of peers, self-regulation is also likely to be playing a role. The small drop in turnout is to be expected as self-regulation is only likely to counter the influence of nonvoting peers if the parents (or other socializing agents) have inculcated a sense of duty to vote. The influence of parents who have treated voting as a matter of weighing the costs and benefits is less likely to withstand peer pressure.

OPEN IN VIEWER

Table 4.

The Estimated Impact of Parental Voting on Their Adult Child’s Turnout According to the Child’s Living Arrangement.

Lives with neither parent
Only father voted	9.4 (0.01)***
Only mother voted	14.6 (0.01)***
Both parents voted	27.7 (0.01)***
Lives with father
Only father voted	30.0 (0.04)***
Only mother voted	12.1 (0.05)*
Both parents voted	40.2 (0.04)***
Lives with mother
Only father voted	7.4 (0.03)**
Only mother voted	36.5 (0.03)***
Both parents voted	44.0 (0.02)***
Lives with both parents
Only father voted	22.9 (0.02)***
Only mother voted	32.7 (0.02)***
Both parents voted	49.2 (0.01)***

Column entries indicate the estimated change in the child’s probability of voting compared with the reference group of “neither parent voted,” estimated on the basis of logistic regression models, with the same controls as Table 3, along with interactions between living with parent(s) and parental voting. Robust standard errors, clustered by family identification number, are shown in parentheses.

*

p < .05, **p < .01, ***p < .001.

The marginal effect of having parents who both vote (compared with having neither parent vote) is much smaller when the adult child has left home. This is because estimated turnout when neither parent votes (32.7%) is significantly higher when the adult child no longer lives with the parents, compared with the adult child who continues to live with the nonvoting parents (16.5%). This underlines an important point. When young adults leave home, they may come under the influence of peers who are less likely to vote than their parents, but if their parents do not vote, having at least some peers who do vote may partially offset the influence of nonvoting parents by encouraging the young adults to vote.

A quarter of the young adults lived with both parents, but 2 percent lived with their father and 6.8 percent lived with their mother. The effect of the voting parent is especially weak when the adult child lives with the parent who does not vote. This is the case whether the nonvoting parent is the mother or the father. Indeed, when the analyses were repeated on split-half samples, in one of the split halves, the effect of having a father who votes was not significantly different from having nonvoting parents if the young adults lived with their mother. The same holds for the effect of having a mother who votes if the young adult was living with the father. This pattern of effects makes it less plausible that the impact of parental voting is an artifact of selection biases as it is very much the pattern we would expect if the adult child is being influenced by the parent, as social learning theory predicts.

Social learning theory gains further support when we look at the effects of parental voting by age. To test this, we added interactions between age and parental voting while controlling for living arrangements. Figure 1 tracks estimated turnout from ages 18 to 30 according to parental turnout, based on this model. The effect of having parents who both vote clearly persists. There is only a slight decline in turnout. There is even clearer evidence of persistence when only the mother votes. However, when only the father votes, there is a significant increase in turnout between ages 18 and 30. The sizable estimated turnout gap at age 18, depending on whether the father votes or the mother votes, reinforces our earlier conclusion about their relative influence. Having a voting mother clearly provides a significant head start. However, by age 30, turnout is almost the same whether it is the mother or the father who votes. This pattern speaks to the importance of life experiences that can make up for the lack of a civic boost that comes from having parents who vote. This becomes even clearer when we look at the doubling of turnout among the adult children of nonvoting parents. This is consistent with Plutzer’s (2002, 48) finding that

parents who said that they did not vote leave their offspring substantially behind in the acquisition of the habit of voting. However . . . this disadvantage dissipates over time.

OPEN IN VIEWER

Figure 1.

The estimated probability of voting according to the child’s age (95% confidence intervals).

Note, though, that at age 30, there is still a significant estimated gap in turnout between young adults whose parents voted and those who did not. We should be cautious, though, about inferring life cycle effects from cross-sectional data.

Conclusion

Our results suggest that status transmission theory may have limited applicability in a country like Finland where children are more likely than their American counterparts to achieve a higher level of education than their parents. Parental education does matter, but only a relatively small portion of its impact on turnout is mediated through the adult child’s education. Parental voting appears to be the more important mediator. The importance of the parental example is further demonstrated by the strong connection between the turnout of parents and their offspring. What is really striking is that this effect is only slightly attenuated by controlling for the parents’ and child’s education and income. Equally striking is the persistence of this link after adult children have left home. Even when young adults are approaching their thirties, the effect of having parents who both vote is only slightly diminished. These findings lend weight to social learning theory. Further support comes from the fact that parental influence is weaker when only one parent votes, which is to be expected given the importance of consistent cues in social learning.

The finding that having a mother who votes is more consequential for young women than having a father who votes is consistent with the emphasis that social learning theory puts on observer-model similarity. However, the father–son link in turnout was weaker than the mother–son link. The fact that the mother’s influence apparently outweighs the father’s is intriguing from the perspective of social learning theory. Observational learning can be based on verbal instruction, which raises questions about the role of communication style. Shulman and DeAndrea (2014) reported that parent–child similarity in party identification and ideology is fostered when parents encourage the open expression of views. They found that mothers are more likely to adopt this style of communication. It may well be that talking about why voting matters is more effective than simply telling children they should vote. Mothers are also more important in terms of transmitting expressive as opposed to instrumental orientations (Finley and Schwartz 2006). Seeing voting as an expression of civic duty (Blais 2000) or as a way of expressing political views and preferences (Schuessler 2000) may be more motivating than viewing it instrumentally as a matter of weighing expected costs and benefits. This is certainly a line of inquiry for future research.

We should, of course, acknowledge the limitations of using cross-sectional data to address questions about intergenerational transmission. A shared social milieu, the local political context and exposure to the same mobilization campaigns could also account for the strong parent–child link in turnout that we have observed. Moreover, the causal arrow could run the other way (McDevitt and Chaffee 2002; Shulman and DeAndrea 2014): a young adult who is excited about the prospect of voting for the first time might influence his or her parents to vote. To make causal inferences with more confidence, we would need data on parental voting during their adult children’s childhood, but such data are not available for the sample used in this study. However, it seems reasonable to assume that many of the parents who voted in 1999 also voted in previous elections, given the evidence that voting is habit-forming (Aldrich, Montgomery, and Wood 2011; Cutts, Fieldhouse, and John 2009; Green and Shachar 2000; Schlozman, Verba, and Brady 2012).

An empirical association between the turnout of parents and their adult children could also reflect genetic inheritance. Recent work suggests that there may be a genetic predisposition to participate in politics (Deppe et al. 2013; Fowler, Baker, and Dawes 2008; Fowler and Dawes 2008; Hatemi et al. 2007; but see Charney and English 2013). Using twin-study data, Fowler, Baker, and Dawes (2008) have shown that genetic differences account for a large and significant amount (as much as 50%) of the variation in voting. Given that biological parents and their children share about 50 percent of their genes, the predisposition to vote may well be heritable, which might help to explain the strong parent–child link in voting that we have observed. We cannot rule out this possibility with the data at hand. There are twins in the data set but there is no way of knowing whether they are identical twins (who share 100% of their genes) or fraternal twins (who share only 50% of their genes). However, we can point to evidence that suggests that genetic inheritance may only be part of the story. Notably, genes cannot explain why the father’s influence is much weaker than the mother’s. Social learning is not incompatible with a genetic explanation: researchers have been at pains to emphasize that environmental factors may condition the influence of genetic inheritance. Our results suggest that social learning within the family may be an important environmental influence.

Despite the limitations, our data are much more reliable than the survey data that are typically used in studies of political socialization and they have enabled us to address a number of important unanswered questions about parental influence. Administrative data have a high degree of accuracy and avoid the problems of faulty recall and social desirability that plague self-reports (Brady, Schlozman, and Verba 2015). This enables us to be much more confident of the empirical association between parental voting and the turnout of their adult children.

The impact of parental voting on young adults’ turnout points to the importance of exploring possible changes in the character of the parental example. There is ample evidence of generational differences in turnout with each subsequent cohort voting at lower levels than previous cohorts (e.g., Bhatti and Hansen 2012b; Blais et al. 2004; Konzelmann, Wagner, and Rattinger 2012; Lyons and Alexander 2000; Wass 2007b). The generational gap in turnout is narrowed when young adults are encouraged to vote by their parents (Wass 2007a), but fewer young adults today actually have parents who vote. Given the substantial effect of having parents who vote, declining parental voting may partly account for declining voting among their children.

Our findings also have important implications for the normative problem of unequal participation. Parental education clearly has an impact on the turnout of young adults: the lower the parents’ level of educational attainment, the less likely their adult children are to vote. One reason is that low-education parents are less likely to vote. The children of university-educated parents, by contrast, enjoy a substantial head start because their parents are more likely to be voters. Having nonvoting parents, though, does not necessarily leave young adults at a permanent disadvantage. With time, it seems, the turnout gap narrows. This points to the role that extra-familial motivational factors and mobilization by voluntary associations, religious organizations, informal social networks, and political organizations can play in at least partially offsetting inherited political disadvantage. However, in countries where low educational attainment is more likely to be passed on from one generation to the next, there may be a more permanent bias in turnout.