Social Network Effects

There is a rich body of literature that documents how one's social context, including networks, discussion and diversity, influences political behaviour. Social context has been found to affect political participation and engagement (Buton et al., Reference Buton, Lemercier and Mariot2012; Harell et al., Reference Harell, Stolle and Quintelier2009; Ikeda and Richey, Reference Ikeda and Richey2005; Kenny, Reference Kenny1992; Knoke, Reference Knoke1990; Lake and Huckfeldt, Reference Due Lake and Huckfeldt1998; Leighley, Reference Leighley1990; McClurg, Reference McClurg2003, Reference McClurg2006a, Reference McClurg2006b; Pattie and Johnston, Reference Pattie and Johnston2009); vote choice (Gidengil et al., Reference Gidengil, Harell and Erickson2007; Huckfeldt and Sprague, Reference Huckfeldt and Sprague1991; Sokhey and McClurg, Reference Sokhey and McClurg2012; Zuckerman et al., Reference Zuckerman, Valentino and Zuckerman1994) and partisan development (Campbell et al., Reference Campbell, Converse, Miller and Stokes1960; Jennings and Niemi, Reference Jennings and Niemi1981; Zuckerman and Kroh, Reference Zuckerman and Kroh2006; Zuckerman et al., Reference Zuckerman, Dasović and Fitzgerald2007).

In each case, the primary mechanism at work is discussion. Through discussion, people become aware of and informed about political issues. Being informed and holding political opinions can in turn lead to a variety of types of political activities.

However, political discussion is not always a neutral enterprise. Discussing political issues with others can and likely will introduce an individual to conflicting attitudes and opinions. What are the effects on political engagement of having a heterogeneous discussion network? To date, the findings are mixed. On the one hand, some researchers have found that political engagement can increase in the presence of disagreement. For example, the work of Cox and Munger (Reference Cox and Munger1989) and that of Blais (Reference Blais2000) suggest that greater political competition increases the closeness of elections, which in turn spurs turnout. Scheufele and colleagues (Reference Scheufele, Nisbet, Brossard and Nisbet2004) find that political participation and hard news media use increase with social network heterogeneity, and McLeod and colleagues (Reference McLeod, Scheufele, Moy, Horowitz, Lance Holbert, Zhang, Zubric and Zubric1999) report positive effects on participation in public forums. Leighley (Reference Leighley1990: 466) finds that discussant conflict stimulates different forms of political participation and Pattie and Johnston (Reference Pattie and Johnston2009) also find that political disagreement can be motivating for many forms of political engagement.

On the other hand, there is also evidence that conflict in political discussion (heterogeneity or disagreement) can lead to a decrease in political engagement. Early research into political behaviour found that cross-pressures delay vote decisions and dampen political enthusiasm and engagement (Campbell et al., Reference Campbell, Converse, Miller and Stokes1960; Lazarsfeld et al., Reference Lazarsfeld, Berelson and Gaudet1944). More recently, Mutz (Reference Mutz2002, Reference Mutz2006) finds that ambivalence towards politics increases when discussion is conflictual and argues that social accountability or conflict avoidance can lead to a withdrawal from political activities in such situations.

Other research provides a more nuanced view of the effects of network heterogeneity. McClurg (Reference McClurg2006a), for example, finds that network conflict has a depressing effect on political activity when people are in the minority of public opinion in their neighbourhood. This is consistent with Noelle-Neumann's “spiral of silence” interpretation (Reference Noelle-Neumann1974) of public opinion. If someone whose opinions are not in line with the majority of the population encounters political disagreement they are less likely to engage due to the perceived lack of social support and fear of isolation. Complementing this work, Nir suggests that conflict within social networks has nonlinear effects (Reference Nir2011). Disagreement in oppositional networks (those in which network members are mostly in conflict with the respondent) demobilizes political engagement. However, respondents with mixed networks, where there is competition between different viewpoints but the respondent has at least some support, tend to be mobilized toward engaging in politics.

Fitzgerald and Curtis argue that disagreement in family environments should have a different effect, as social accountability concerns should be weaker (Reference Fitzgerald and Amber Curtis2012: 131). In their research, partisan conflict between parents has an encouraging effect on political interest when the electoral context is less competitive. Thus, how and whether one experiences social pressure may be the key to understanding the effects of network conflict on political behaviour.

Beyond the context in which heterogeneous discussion occurs, some research also points to individual-level variation. Nir (Reference Nir2005), for example, finds that being exposed to cross-pressures can help certain people (those who are less ambivalent) make early vote decisions. Also supporting the possibility of individual-level variation is the work of Djupe and colleagues (Reference Djupe, Sokhey and Gilbert2007) who find that practising civic skills in church is influenced by similarity to the church group for women, but not men. Scheufele's work (Reference Scheufele1999) also finds that demographics and personal predispositions are related to public opinion expression.Footnote ¹

Are Social Network Effects Different for Young People?

Despite the considerable breadth of the research discussed above, studies of social network effects have yet to explore whether the relationships between social network discussion and disagreement and political engagement differ by age cohort. To be sure, certain research studies have focused on young people, such as the work of Harrell and colleagues (Reference Harell, Stolle and Quintelier2009) and Quintelier and colleagues (Reference Quintelier, Stolle and Harell2012), using surveys of young people in Belgium and Canada, and research done by Dostie-Goulet (Reference Dostie-Goulet2009). The findings in these studies, however, provide only incomplete evidence as no direct comparisons are made to other age groups and they do not theorize about how the effects of political discussion and conflict may differ between young people and older adults. Thus, this article aims to build upon existing research by explicitly comparing the relationships between young people's political engagement and network discussion and conflict to those of older adults.

Of course, our theoretical starting point is not a tabula rasa. Extant literature provides reason to think that age might be a factor in the effects of social network discussion and conflict. For example, Howe (Reference Howe2010) demonstrates that political knowledge has a more dramatic positive effect on turnout for young people than older adults. As one of the functions of political discussion is to provide information, this finding suggests that discussion in social networks may be more beneficial for encouraging political engagement in young people compared to older adults. Young people tend to have lower levels of knowledge to begin with, so they simply may have more to gain from receiving information.

Findings for social network conflict also suggest that young people may be more positively affected by their social networks than older adults. Quintelier and colleagues (Reference Quintelier, Stolle and Harell2012) and Harrell and colleagues (Reference Harell, Stolle and Quintelier2009) both find that political diversity has a motivating effect on the political activity of young people, contrary to the depressing effect found for the general population by some others. The work of Fitzgerald and Curtis (Reference Fitzgerald and Amber Curtis2012) is also relevant as the respondents in their dataset have a mean age under 30 and they report neutral or motivating effects of parental partisan conflict (but never negative). It is therefore possible that the social network experiences of young people may be different from those of older adults. If the social accountability mechanism, or the fear of isolation due to unpopular political expression, is weaker for young people, then the potential for network heterogeneity or conflict to have a depressing effect on political engagement will be much less. Given lower expressions of political duty and efficacy among younger adults (Blais and Rubenson, Reference Blais and Rubenson2013), being differentially affected by political heterogeneity is not hard to imagine.

The analyses in this article are designed to consider the relationship between social networks and political engagement for young people, in comparison with the effects among older adults.

Hypotheses

We consider a number of different hypotheses in this article. First, we are interested in any differences in the nature of the engagement-political discussion relationship in each age cohort. Given the discussion above, we expect that political discussion will have a positive relationship with political engagement and that the relationship will be stronger for young people.

We also expect that social network heterogeneity will affect political engagement positively and more strongly for young people, given the research findings for young people and in spite of the mixed findings for adults.

Our dependent variable, political engagement, is a broad term. There are many different ways that individuals may engage in the political realm, from being interested to voting to working for a party to giving a monetary contribution. Work by Pattie and Johnston (Reference Pattie and Johnston2009) suggests that the specific type of political activity being considered matters for the effects of disagreement. As noted earlier, there is also research that indicates young people participate differently in politics than older adults (O'Neill, Reference O'Neill2007; Turcotte, Reference Turcotte2015). Dalton (Reference Dalton2015) takes this a step further and suggests that there are different views of what makes a good citizen. Young people's political participation is not the same as that of older adults because it reflects a different conception of good citizenship.

In order to investigate whether social network discussion and conflict affect youth political engagement, then, we need to be sure that we evaluate a number of different dependent variables. We conceptualize political engagement as a hierarchy of increasingly public and committed activities within the political sphere. We first consider simple expressions of political interest, as it requires little effort on behalf of the respondent and is not necessarily partisan. Our second measure of political engagement is an expression of party identification that reflects an attachment to a particular party and necessarily conveys engagement with the political sphere and some of the debates within it. Finally, we consider political engagement through different forms of political activity. We make a distinction between public and partisan activities (for instance, displaying a campaign lawn sign or attending a political rally) and those that are not (anonymously posting a comment online or voting). Our rationale is that there is a qualitative difference in the degree of commitment between those activities that are non-partisan and those that are partisan, and the public component follows from Campbell's model of peer influence (discussed below) in which he expects influence to be stronger when the attitude is “‘visible,’ that is, it should have a behavioral manifestation” (Reference Campbell1980: 326). Further, differentiating types of participation also allows us to investigate some of the considerations in O'Neill (Reference O'Neill2007), including that young people participate differently and take part in individual activities more than group ones. While we might suspect that encouragement to participate is most likely to create a difference in the more public and partisan activities, we are mindful that if young people do not prefer such activities, then mobilization may lead to more individualized and less traditional activities. Our expectations are therefore that there will be variation between younger Canadians and older adults and that these will vary by the type of engagement:

Finally, we address one additional potential source of variation: the discussion group itself. Campbell (Reference Campbell1980) develops a theory of group influence in which the importance of the group and attitude being influenced to the individual, as well as the visibility and importance of the attitude to the peer group, strengthens the influence of the group. He develops this theory based upon ideas from Festinger and colleagues (Reference Festinger, Schacter and Bach1950). The key idea is that peer influence is not guaranteed; there must be relevance for the attitude being developed and holding compliant attitudes must matter (and be visible) to the group. Campbell finds that political attitudes are influenced by peers, although less so than (more visible) racial attitudes. He also finds that the amount of influence is positively related to the political involvement of the group (the attitude's importance) for political efficacy and political trust, but not partisanship.

Similarly, if disagreement is expected to affect individuals through the mechanism of social accountability, it follows that in contexts where social accountability pressures are low any dampening effect of heterogeneity on political engagement will also be low. As Fitzgerald and Curtis explain, “In contrast to less intimate settings, social accountability considerations should be lowest within the private context of the family. Discussion among family members should be more frank and less inhibited by norms of social decorum than exchanges with less familiar acquaintances. Within this safer social environment, the informing mechanism of disagreement—which pushes discord's effects in a positive direction—should be enhanced” (Reference Fitzgerald and Amber Curtis2012: 131).

Following this logic, we suspect that there may be differences in the relationships of discussion and disagreement with engagement across the types of social networks. In particular, we look at discussions with family and household members, friends and work colleagues. These types of social networks involve different degrees of discussant choice, familiarity and permanence.

We also have reason to suspect variation by age. According to Howe (Reference Howe2010), there is a greater period of adolescence in modern society, which increases the influence of peers on individuals. Given the importance of peers to young adults there may be differences in the degree of social accountability pressures in friend networks across age cohorts. However, following Campbell (Reference Campbell1980), if politics is not considered to be a domain of importance to one's friends, then such influence may be muted.

We expect household discussion to affect both young and older adults similarly, as socialization begins at home and there is evidence that the political activity of other household members is influential (Buton et al., Reference Buton, Lemercier and Mariot2012; Fieldhouse and Cutts, Reference Fieldhouse and Cutts2012). For work networks, from Mutz and Mondak we know that “the workplace is the leading context for exposure to cross-cutting political discourse” (Reference Mutz and Mondak2006: 151) and that such exposure is related to political tolerance and awareness of rationales for opposing views. In terms of the sheer number of hours spent with such colleagues, the impact of discussion and disagreement may be strongest for older adults, for whom such contact is likely to be more intense. However, if young people consider their work colleagues to be peers, then their influence may be quite strong as well.

Given the variety of effects indicated in the literature, we do not have firm expectations for which networks will affect young people more or less than older adults. We do, however, believe that it is an important question to explore. Therefore, our hypothesis is quite general:

Data and Methodology

To evaluate these hypotheses, we use a dataset collected through an online survey of 2620 Canadians from February 1 to 26, 2013, by Harris/Decima. The survey was designed to be nationally representative of region, gender and language in Canada and includes an oversample of 1002 young adults (30 and under). All analyses for this project use demographic and propensity weights. This dataset enables us to test, using identical questions and methods, whether the relationships between discussion and disagreement and engagement differ by age cohort. Although we cannot demonstrate causality, we can provide evidence suggestive of the direction and intensity of the relationship as a first (and important) step in better understanding the importance of social networks for the politicization of young people.

Our dependent variables for this article are five different measures of political engagement. The first, political interest, is a general measure of cognitive engagement with and attention to the political sphere, which has been used in other studies (see, for example, Fitzgerald and Curtis, Reference Fitzgerald and Amber Curtis2012). Our measure is built from a question that asks “How interested are you in FEDERAL politics? (very interested, somewhat interested, not very interested, not at all interested, don't know, prefer not to say).” Our second dependent variable is partisanship. Rather than focus on party identification as an indicator of support for a specific party, we conceptualize it as a form of political engagement. That is, we contend that if one identifies with a party then he or she has chosen to situate him or herself in the political arena and has chosen a “side,” so to speak. To this end, we are not concerned with the direction of one's identity so much as whether or not the individual reports having one. The partisanship measure included in our survey, commonly used in comparative work on partisanship, presents an initial question before providing specific party options. This “closeness to a party” measure asks: “Do you usually think of yourself as close to any particular FEDERAL political party? (yes, no, don't know, prefer not to say).”

The final dependent variables are measures of political activity. Respondents in our dataset were asked to “indicate how often you participate in the following POLITICAL activities? (regularly, sometimes, rarely, never, don't know, prefer not to say) Vote in elections, Give money to a candidate or political party, Try to convince friends to vote for or against a particular candidate or political party, Work for a candidate or political party during a campaign, Put campaign signs in your yard or window, Attend meetings or election rallies for a candidate or political party, Leave comments on political blogs or online news articles about politics, Use social media (including Twitter and Facebook) to comment on politics, including “liking” and joining groups.” Our first measure of political activity is voting. We then divided the remaining activities to create two separate measures of political activity: partisan campaigning (give money to a candidate or party, work for a candidate or party, put campaign signs in yard or window, attend meetings or ralliesFootnote ²) and online engagement activities (leave/post comments online, use social media to comment). We make this distinction mindful of the argument of Quintelier and colleagues (Reference Quintelier, Stolle and Harell2012) that the mobilizing effects of network diversity will overcome negative effects for non-partisan activities. Our five political engagement variables range in the degree of commitment to the political arena, as discussed earlier, from political interest to non-partisan activity to party identification to partisan activity.

Our independent variables of network discussion and conflict come from questions that probe relationships with close friends, work colleagues, and household members.Footnote ³ For each discussion network, we have measures from the following question: “In general, among all of your close friends/work colleagues/household members, how often do you discuss politics? (often; sometimes; rarely; never; don't know; prefer not to say).” We code this variable 0 to 1 for increasing frequency of discussion. Our measure of disagreement asks: ‘When you discuss politics with people in each of these groups [close friends/work colleagues/household members], how often do you tend to disagree? (often, sometimes, rarely, never, don't know, prefer not to say).Footnote ⁴ We code this variable 0 to 1 for increasing frequency of disagreement. We combine responses for each of household, friend and work colleague networks into a single measure of disagreement for each of the disagreement types to investigate hypotheses 1–3, and use the disaggregated variables for H4. While we recognize that each of these measures is subjective and depends upon the respondent's own recollections, we follow Mutz in asserting that “it makes…theoretical sense to argue that respondents will experience cross-pressures to the extent that they recognize that their network members hold differing political views” (Reference Mutz2002: 843).

Discussion, Disagreement and Engagement

We begin our empirical consideration of the research questions of this article by exploring the distributions of the key dependent and independent variables by age cohort. We first consider distributions of discussion and disagreement and then present the various indicators of political engagement. Do young people engage in less political discussion, experience less political conflict and are they less engaged than older respondents? Table 1 presents descriptive statistics that bear on these questions.Footnote ⁵ The mean cumulative measure of discussion across all three social networks for young people is 0.43 (on a 0-1 scale). By contrast, the mean score for the same measure of discussion among adults over the age of 30 is 0.49. This 6-point difference is significant at p < 0.001. Similarly, the average score for network disagreement among adults under the age of 31 is 0.41 (again, on a 0-1 scale) and the comparable value for older respondents is 0.46. This 5-point gap is significant at p < 0.001. Simply put, our data clearly indicate that younger adults discuss politics less and have fewer disagreements about politics than older adults.

Table 1 Discussion, Disagreement and Engagement Types by Age Groups

Bold type indicates difference of means between young people and older adults is significant at p < 0.001.

Table 1 also presents a comparison of our measures of political engagement by age cohort. On average, older cohorts (above the age of 30) are more interested in politics (by 12 points), more likely to hold a partisan identity (by 8 points), more likely to vote (by 12 points) and are more likely to engage in party campaigning (by 5 points) than respondents under the age of 31. Each of these differences is statistically significant at p < 0.001. By contrast, younger respondents are more likely to participate in forms of online political engagement (by 13 points) as compared to respondents over the age of 30. This difference is statistically significant at p < 0.001. These means reflect the predominant narrative that younger voters are somewhat less engaged than older voters. However, the online engagement result suggests that opportunities for engagement through technological advances have been largely picked up by younger citizens.

Another preliminary question is whether young people simply have fewer overtly political social networks than older adults. To provide a simple indicator of this, we consider the number of reported “don't knows” for the partisan leanings of their discussion network, by network type and age cohort. For the sample as a whole, 20.8 per cent of respondents report not knowing the partisan identifications of their friend network, 39.9 per cent indicate not knowing this within their work network and 17.4 per cent for their household discussion network. There are no substantive differences between young and older adults in the incidence of not knowing partisan identifications for both the friend and household discussion networks. However, there does appear to be a significant difference based on age within the work network. In particular, older respondents are more likely than younger adults to report not knowing the partisanship of their work network.

These descriptive data suggest that young people engage in less political discussion, report less conflict and do not have markedly less politicized social groups except for the work network. Thus, there are some differences and some similarities between the social experiences of young and older adults.

How Does Discussion and Disagreement Affect Political Engagement?

To investigate the relationships between political discussion and disagreement and types of political engagement, we first estimate models using the whole data set—all respondents aged 18 and older—to provide a baseline against which to compare cohort effects. All models control for education, gender and sense of duty to turn outFootnote ⁶ and separate out discussion and disagreement. Table 2 shows the results from these regressions (OLS for each form of engagement except turnout where we utilize a logistic regression).

Table 2 Baseline Models: Relationship of Network Discussion and Disagreement with Political Engagement

* p < .1; ** p < .05; *** p < .01

In the first instance, it can be observed that in each model the coefficient for political discussion is significant and in the expected positive direction: more political discussion within the social network correlates with greater amounts of political engagement across our five indicators of engagement. We also find that the coefficient for network disagreement is only significant for turnout. Those who experience disagreement about politics within their network are more likely to turn out. This finding provides support for the idea that disagreement mobilizes political engagement. However, although insignificant, the coefficients for three of the remaining four measures of engagement are negative. This suggests that the mixed findings in the literature may be realized elsewhere as well.

We also expect that the effects of discussion and disagreement may increase based on the substantive nature of these forms of engagement (H3). Based upon average marginal effects (displayed in Table 3), the relationship with political discussion is strongest for political interest and partisan engagement and smallest for non-partisan engagement, which does not follow the hypothesized pattern.Footnote ⁷ By contrast, the only significant effect for political disagreement is found for turnout. Experiencing political opposition within one's discussion networks is not significantly related to expressing interest in politics or becoming a partisan but is related to turning out. However, there is also no significant relationship for the other forms of engagement, so our full expectation is not supported.

Table 3 Average Marginal Effects of Baseline Models: Network Discussion and Disagreement with Political Engagement

* p < .1; ** p < .05; *** p < .01

Are There Cohort Differences?

To assess the effects of discussion and conflict on younger Canadians’ political activity in a multivariate context we ran regressions for each of our dependent variables (OLS for all but turnout, which was a logistic regression) with our key independent variables and interaction terms. We included a variable to indicate those aged 18 to 30 and then interacted that variable with our two measures of discussion and disagreement. We also included education, gender and sense of duty to turn out as control variables, as above.Footnote ⁸

H1 suggests that social network discussion will have a greater positive relationship on engagement among young adults as compared to older adults. The results in Table 4 provide support for this assertion. In the first instance, the core relationship of the discussion term remains positive and significant suggesting that more political discussion among older Canadians has the positive, mobilizing relationship we expect. Beyond this, for the models with the dependent variables of political interest, party campaigning and online engagement, the mobilization effect of political discussion is significantly stronger among our respondents under the age of 31 in comparison to those over the age of 30. This is an important finding as it shows that independent of the levels of political discussion by cohort, the positive relationship between political engagement and network political discussion is stronger among younger voters. This suggests that there is something unique about being a younger person who discusses politics.

Table 4 Effects of Discussion and Disagreement on Engagement with Youth Interactions

* p < .1; ** p < .05; *** p < .01

We next consider whether the political disagreement relationship is greater among older adults than young adults, and whether it is in a different direction (H2). The results of our baseline analyses showed that disagreement had a positive relationship with turnout. The results in Table 4 support this result, as the base term remains significant and positive in that regression and the interaction term is not significant. There is no significant difference in the relationship of disagreement and turnout among younger Canadians. However, there is one interaction term that is statistically significant: younger respondents who experience more network disagreement are less likely to be involved in party campaigning. Of the four other (insignificant) coefficients, three are also in the negative direction. This is an interesting finding because it speaks to the debates and conflicting findings about the effects of disagreement, and suggests that there may indeed be a demobilizing effect of disagreement, as Mutz (Reference Mutz2002) suggests and contra Fitzgerald and Curtis (Reference Fitzgerald and Amber Curtis2012).

Finally, we take up consideration of H3, that the effects of discussion and disagreement will become greater as the type of engagement becomes more public and partisan. In order to evaluate this, we report the average marginal effects of each of our political discussion and disagreement interactions from Table 4 (see Table 5).

Table 5 Average Marginal Effects of Baseline Models: Network Discussion and Disagreement Interactions with Political Engagement

* p < .1; ** p < .05; *** p < .01

By delving into the type of engagement, the value and nuance of network discussion becomes clearer. The marginal effect size for the young people*discussion interaction term is comparable in magnitude for each of political interest and party campaigning, suggesting that the relationship between political engagement and political discussion among young people is stronger than for older people but similar across these forms of political engagement. By contrast, the average marginal effects for the youth*discussion term for online engagement is about two times larger. This suggests that young people who discuss politics are much more likely to be engaged in online comments and social media interaction, and reflects our earlier descriptive results.

In addition, where significant, the effect of political discussion among young people somewhat follows the pattern predicted as the significant relationships for the interaction term is smallest for political interest (0.21), somewhat larger for party campaigning (0.28) and largest for online engagement (0.53). There is no significant effect, however, for party identification or turnout.

We find little evidence for H3 with respect to disagreement as only party campaigning is affected by the youth*disagreement interaction term. Reporting more disagreement in one's networks is related to lower levels of party campaigning, but it has no significant relationship with any other type of engagement. We therefore see no pattern with respect to partisanship or public visibility.

Disaggregating Networks

In the above analyses we considered the aggregate social networks of individuals, including friends, work colleagues and household members. The results suggest that the relationship between political discussion and engagement varies across age cohorts for four of our five measures. However, the effects of disagreement on engagement only varied by age for one measure of engagement, party campaigning. These results are somewhat disappointing given our expectations. However, it is possible that there are differences between younger and older adults across types of discussion networks. In this section, we evaluate that possibility, in line with our fourth hypothesis, by running the same multivariate models as presented in Table 4 separately for each type of discussion network. The direction and significance of the average marginal effects from these analyses are summarized in Table 6.

Table 6 Average Marginal Effects of Youth and Network Discussion/Disagreement Interactions by Network Type

* p < .1; ** p < .05; *** p < .01

The results suggest some interesting dynamics across types of discussion networks. First, the overall enhanced positive relationship for discussion among young people is most prevalent within the household discussion network (4 of 5 engagement measures) and least for friend and work networks (2 of 5 engagement measures). For household discussion networks, more discussion has a stronger relationship with young people's political interest, their party identification, incidence of party campaigning and online engagement. Based on a comparison of the average marginal effects between these network types, it would appear that discussion within the work network has the greatest effect on young people's interest in politics in comparison to household networks (0.16 vs. 0.12). This said, household discussion exhibits the most consistent effects among the three network types across these measures of engagement and has the greatest effects on online engagement: 0.42 versus 0.37 (friends) and 0.22 (work). In sum, the most consistent relationships between discussion and engagement among young adults are found when that discussion occurs in a household network.

Although the weaker findings for friend networks are contrary to our expectations, the strong household effects are not unexpected given existing research. We also find the effects of discussion in work networks particularly interesting. Whereas the literature suggests friends might have particularly strong influence over young people, there is no study of work networks. The effect on political interest, specifically, is in line with the work of Mutz and Mondak (Reference Mutz and Mondak2006), such that exposure to cross-cutting information (more common in work environments) is expected to increase political interest. This result may be particularly evident given that young people report more politicized work networks than older adults (Table 1). We can speculate that the differential ability to choose discussion partners in work networks compared to friend networks may be related to the type of political engagement that is related to each type of discussion (party campaigning for friend discussion, political interest for work discussion). It may also relate to the heterogeneity of the individuals in the network, such that one's work environment is likely to bring young people into contact with older individuals with greater and different life experiences than one's friend network. However, investigating these ideas is beyond the scope of our current study. Overall, our results suggest that political discussion in any forum appears to have some positive relationship with engagement for young people.

Regarding disagreement, the results are notably limited. When disaggregated by network type, there is only one instance where network disagreement has a stronger effect among young adults as compared to older adults. The incidence of online engagement among young adults is heightened when they disagree about politics with household members. This finding may indicate that the online engagement activities of young adults are motivated by disagreement in the household, which somewhat echoes Fitzgerald and Curtis's finding (Reference Fitzgerald and Amber Curtis2012) that political interest is increased when parents prefer different parties.

Implications and Future Directions

We began this study with the hope of clarifying how social network discussion and conflict impacts young people in comparison to older adults. Based on the extant literature, we know that political discussion is a motivating factor for political engagement. By contrast, the literature reports mixed findings regarding the effects of disagreement on engagement. With this literature providing our empirical starting point, we considered the core question of this article: are there systematic differences between age cohorts in terms of the relationships of discussion and disagreement with political engagement?

Our findings, while interesting, are mixed. On the positive side, the overall effect of political discussion is confirmed for both age cohorts. Political discussion mobilizes across all of our measures of political engagement. Likewise, our expectation that discussion should have greater effects on the engagement of young people is supported. Political discussion is related to more political engagement, and these effects are often greater for young people. Beyond this, our results for separate networks confirm existing understandings of the importance of household political discussion and lead us to speculate that work environments might be important arenas for inspiring general political interest among young people. Therefore, when seeking to increase engagement among young adults, simply facilitating political discussions is an excellent starting point.

Turning our attention to the effects of network conflict, our results indicate fewer age cohort differences. While we had no specific directional expectations, we did find that political interest is lower when disagreement is experienced by young people, compared to older adults. However, this effect does not appear to be driven by a specific type of network, and online engagement is related to more household disagreement. Thus, disagreement seems to have an inconsistent relationship with political engagement for young people.

The results of this article suggest that political discussion is a relevant factor in engagement for young people. As young people report less discussion and less disagreement than older adults, perhaps providing more forums for discussion among groups of all types is a simple way to increase youth political engagement.