Introduction
Declining voter turnout has been most pronounced among young adults. While antidotes remain elusive, we know that voting and abstention are habit-forming (for example, Blais, Reference Blais2000; Campbell, Reference Campbell2006; Franklin, Reference Franklin2004; Gerber et al., Reference Gerber, Green and Shachar2003; Johnston et al., Reference Johnston, Matthews and Bittner2007; Plutzer, Reference Plutzer2002). Initial turnout decisions made by those aged 18–30, approximately, set a course for their political engagement in adulthood. By implication, the immediate pre-adult years may offer important insight into why young people do or do not vote when they become eligible. Opportunities to examine this segment of the population are rare, because studies tend not to analyze individuals who are under the age of 18.
Our data come from a survey of high school students in Grades 10 to 12 conducted by the Ontario Students' Assembly on Electoral Reform. Our primary research question is whether people in the years immediately prior to the voting age—adolescents—see themselves as future voters. To explain variation in turnout views, particular attention is paid to four types of variables: socio-demographic, socialization, attitudinal and political knowledge. Findings suggest that socialization experiences and some political attitudes have profound impacts on expected political involvement among teenage youth.
Youth and Turnout
In this section, we have two tasks. First, we provide a brief overview of the factors scholars tend to associate with declining youth turnout. Our second goal for this section of the paper is to evaluate collective efforts toward understanding of youth turnout decline given that the voting behaviour literature tends to ignore the pre-adult years.
Scholars estimate that turnout among young Canadians born since 1970 is 20 per cent lower than turnout among baby boomers (born 1945–1959) at the same age (Blais et al., Reference Blais, Gidengil, Nevitte and Nadeau2004: 225). Similar patterns have been reported cross-nationally (for example, Blais, Reference Blais2000; Franklin, Reference Franklin2004). Explanations for declining turnout among the youngest age group have tended to be framed in terms of a debate between life-cycle, generational and/or period effects (and principally between the former two effects). Within this, there are a number of variables known to correlate with the pronounced youth turnout decline: singlehood (Stoker and Jennings, Reference Stoker and Jennings1995); greater mobility (Squire et al., Reference Squire, Wolfinger and Glass1987); lower political knowledge (Gidengil et al., Reference Gidengil, Blais, Nevitte and Nadeau2004; Howe, Reference Howe2006; Milner, Reference Milner2002); reaching political adulthood in an atmosphere of political uncompetitiveness (Franklin et al., Reference Franklin, Lyons and Marsh2004; Johnston et al., Reference Johnston, Matthews and Bittner2007); declining levels of civic duty (Blais, Reference Blais2000); a preference for non-electoral political participation (for example, Norris, Reference Norris2002; but see Gidengil et al., Reference Gidengil, Blais, Nevitte and Nadeau2004; Young and Everitt, Reference Young and Everitt2004); value change associated with a “decline of deference” (Nevitte, Reference Nevitte1996); and, relatedly, greater partisan de-alignment (Dalton and Wattenberg, Reference Dalton and Wattenberg2000).
Singlehood and increased mobility tend to be classified as life-cycle effects. Lower political knowledge, electoral uncompetitiveness,Footnote 1 declining civic duty, preference for unconventional participation, value change and partisan dealignment can be grouped under the heading of generational effects. While lower turnout and other forms of participation decline among youth tend to be linked with pronounced political cynicism, particularly in popular discussion, youth are actually no more cynical than other cohorts. In fact, youth may be less cynical than older generations (for example, Blais et al., Reference Blais, Nadeau, Gidengil and Nevitte2002; Nevitte et al., Reference Nevitte, Blais, Gidengil and Nadeau2000; O'Neill, Reference O'Neill2001; Pammett and LeDuc, Reference Pammett and LeDuc2003). The point here is that there have been a variety of factors linked with declining youth turnout.
Our contribution to the youth turnout decline debate is uncommon in its focus on adolescents. Aside from notable exceptions (for example, Hooghe and Stolle, Reference Hooghe and Stolle2004; Campbell, Reference Campbell2006), work on turnout decline tends to exclude those in the pre-adult years. It is worth questioning how our tendency to study people who have already reached the voting age affects our thinking about youth and participation. As Niemi so astutely observes, “political ideas—like the consumption of cigarettes and hard liquor—do not suddenly begin with one's eighteenth birthday” (Reference Niemi and Knutson1973: 117). Political socialization—how individuals become part of their political communities and acquire attitudes toward political actors, symbols and behaviours—is a process that starts in early childhood and continues into adulthood. Understanding young adults' political behaviour requires more serious attention to the pre-adult years.
It is during the adolescent years that we can identify “empirically founded hints at developmental risks” (Krampen, Reference Krampen2000: 278) that may result in lower political participation in adulthood. For reform minded scholars and policy makers, this means that extending our analytical lens to examine youth (younger than 18), not just young adults (18 and above), holds purchase for devising methods for encouraging turnout. Our article is informed by recent work on the “developmental theory of turnout” (Plutzer, Reference Plutzer2002), for we ascribe to the claim that turnout is a function of both “starting point” and “inertia” (41), that is, the likelihood one will vote in the first eligible election and the tendency for initial turnout decisions to become habitual. Our contribution in this article is to sort out the factors related to the “starting point” —specifically, what types of Canadian adolescents see themselves as future voters—and to draw out the implications for understanding turnout decline among young adults.
The Ontario Students' Assembly Survey
Our data come from a survey conducted by the Ontario Students' Assembly on Electoral Reform,Footnote 2 a process parallel to the Ontario Citizens' Assembly on Electoral Reform. Among other activities, the organizers of the Students' Assembly developed curricula for Ontario secondary school teachers to use for teaching modules on electoral reform to high school students. For each classroom that opted to participate, the unit on electoral reform was meant to culminate in a Classroom Assembly on Electoral Reform. The data we analyze in this article come from a survey that was designed by the authors of this paper and then supplied by the Students' Assembly organizers to high school teachers as part of the in-school unit on electoral reform. For the most part, students completed the questionnaire at the end of the unit on electoral reform. Over 800 high school students representing 21 Ontario ridings completed the survey.Footnote 3
These data have limitations. First, the sample is not a representative cross-section of Ontario high school students. The participation of classes in the survey was at the discretion of individual teachers.Footnote 4 Second, the context in which the survey was conducted may have differed from one classroom to another. While the Students' Assembly organizers provided recommendations for conducting the survey, there was no central oversight to ensure uniformity in survey administration.Footnote 5 As a result, we exercise caution about the extent to which our results can be generalized to the high school population at large.
Our dependent variable for these analyses is anticipated turnout to vote. The survey question used to measure this intention among respondents asked, “If there was a provincial or federal election tomorrow and you were eligible to vote, would you vote?” While general and hypothetical given its forward-looking nature, this question provides a picture of whether adolescent respondents see themselves as future voters. The anticipated turnout question provided three response categories: “yes,” “no” and “don't know.” The “don't know” option was included due to the possibility that students may not have previously considered whether they would vote and, therefore, may have no opinion on the matter. We exclude the “don't know” respondents from our regression estimations, opting to include only those respondents who had clear answers about whether they would vote if they were eligible.Footnote 6
Theory and Methods
To account for expected turnout decisions, we begin with a set of background characteristics: gender (male or female), grade (10, 11, or 12) and religion (Catholic or not).Footnote 7 We expect no difference in anticipated vote turnout between adolescent boys and girls. There have been no gender gaps in turnout for decades either in Canada or in most advanced democracies (for example, Everitt, Reference Everitt1998; Schlozman et al., Reference Schlozman, Burns, Verba and Donahue1995). If there is any gender difference in turnout, girls may be slightly more likely to view themselves as future voters (Hooghe and Stolle, Reference Hooghe and Stolle2004), a finding that is also reflected in the actual turnout of adult populations (for example, Carroll, Reference Carroll2006; Pattie and Johnston, Reference Pattie and Johnston2001). In their analysis of the attitudes toward future participation among fourteen-year-old Americans, Hooghe and Stolle speculate that gender gaps are less likely to occur in the adolescent population because political resources such as “time, money, and cognitive skills” are more equitably distributed among individuals in this age group (Reference Hooghe and Stolle2004: 3).
The survey was administered to students in Grades 10 to 12, so our grade variable is ordinal with one category per grade level. We expect a positive relationship between grade and turnout. Students in more senior grades are quickly approaching political adulthood, and some have already reached eligible voting age.Footnote 8 Therefore, voting may be more relevant to this group. Additionally, senior students will have completed the Grade 10 civics course that is compulsory in Ontario. These students may possess both stronger attitudes of civic duty, as well as the knowledge of how to vote due to this learning experience in the classroom.
Our last socio-demographic variable is a dichotomous variable representing whether a respondent is Catholic or not. The growing secularization of society may render the religious component of Catholicism (or any other religion, for that matter) less relevant for today's youth compared to youth of past generations, yet there appears to be something influential in the cultural legacy of Catholicism that makes Catholics more deferential to hierarchy and authority than their Protestant and other counterparts, even today. In addition, there is also evidence that Catholics are particularly collectivist or communal compared to their Protestant counterparts, an insight noted famously by Weber (Reference Weber1930) and which continues to comport with current realities (for example, Cohen and Hill, Reference Cohen and Hill2007; Cohen et al., Reference Cohen, Hall, Koenig and Meador2005; van Kersbergen, Reference van Kersbergen, Kitschelt, Lange, Marks and Stephens1999).Footnote 9 Applied to turnout, the implication is that Catholics may be more likely to vote (or see themselves as future voters) than Protestants because of a greater sense of duty to fellow citizens to participate in collective political decision making. After all, “the core element of collectivism is the assumption that groups bind and mutually obligate individuals” (Oyserman et al., Reference Oyserman, Coon and Kemmelmeier2002: 5).Footnote 10 In this sense, one might say that Catholicism is fertile ground for the development of civic duty, which tends to have a powerful positive effect on turnout (Blais, Reference Blais2000).
We expect socialization agents (including family, peers and school) to have strong effects on adolescents' expectations about their future turnout (for example, Hyman, Reference Hyman1969; Krampen, Reference Krampen2000).Footnote 11 In line with the bulk of scholarship on socialization, we expect family to have a significant influence on whether students see themselves as future voters. Presumably, the actions, attitudes and conversations within students' home environments have important effects on their views about politics and on their expectations about future political involvement. We operationalize family political socialization experiences using a variable that asks respondents the extent to which politics is discussed in the home, which is relatively common in the literature (for example, Meadowcroft, Reference Meadowcroft1986; Valentino and Sears, Reference Valentino and Sears1998; Westholm, Reference Westholm1999). This is a four-category ordinal variable that ranges from “never” to “regularly.” Greater exposure to political discussion at home may instill in young people the idea that politics is relevant to their lives (for example, Meadowcroft, Reference Meadowcroft1986; Westholm, Reference Westholm1999). Political discussion with parents may also stimulate political interest and generate new knowledge for young people, both of which create affective and cognitive engagement with the political system, consistent predictors of turnout in the literature on the adult population. Even when political discussion is initiated by children, not parents—a possibility rarely acknowledged in the political socialization literature—parents' creation of a home environment where discussion and open debate are encouraged and where children's political viewpoints are not trivialized are important components of the socialization process for adolescents.
This brings us to an important point. We do not conceptualize political discussion or political socialization, for that matter, in singular terms. Recent work emphasizes that political socialization is an active process that includes young people's participation (for example, McDevitt and Chaffee, Reference McDevitt and Chaffee2002). Consequently, we cannot assume that all family political discussion is initiated by parents. Children can and do initiate political discussion, sometimes as a consequence of exposure to politics in news media or in school-based civics courses (McDevitt and Chaffee, Reference McDevitt and Chaffee2002, Reference McDevitt and Chaffee2000). Socialization is a multi-faceted, multi-directional process, and we are cognizant of the fact that parsimonious indicators necessarily simplify the complexity of the process. In the end, we expect that respondents who report more frequent home-based political discussion will, for a variety of reasons, be more likely to see themselves as future voters than adolescents who are rarely exposed to (or feel uncomfortable initiating) political conversation in the home.
Our peer socialization variable is operationalized using a survey question that asks respondents about the nature of their group involvements. Respondents were asked about their participation in a variety of groups from student council to ethnic associations to organized sports. We are concerned only with those involvements where peer interaction can be assumed to occur regularly. Thus, we make an analytical distinction between four involvements that are focused quite obviously on adolescent peer interaction (sports/recreation involvements; community youth group; student council; other school group) and five involvements that are not (music/arts/literary groups; religious groups; environmental groups; ethnic associations; other community groups). In other words, with this latter group of involvements, we have no way of knowing whether respondents' interactions in these groups are primarily with peers or with adults. The peer involvement variable is coded dichotomously: respondents who are involved in one or more of the four peer-interaction groups receive a score of 1, and those who are not involved in at least one of these peer groups receive scores of 0.
While group involvement, even in seemingly apolitical groups, has important linkages with political participation for all age groups (for example, Erickson and Nosanchuk, Reference Erickson and Nosanchuk1990; Putnam, Reference Putnam2000), our interest in group memberships is narrower. We focus on group involvement as an opportunity for organized and goal-oriented peer socialization. We expect that adolescents who are particularly involved in groups with their peers are more likely to see themselves as future voters. Indeed, these active teens are likely embedded within peer cultures that emphasize achievement and motivation, and they may also have greater confidence and efficacy gleaned from their mobilization with other adolescents. Adolescents who are involved in peer-based groups have also probably acquired skills that can be transferred to the political sphere when they reach adulthood (for example, Hanks, Reference Hanks1981; Verba et al., Reference Verba, Schlozman and Brady1995).Footnote 12
Another important socialization influence is school political learning experiences. While high schools in Ontario have a mandatory Grade 10 civics course, other courses such as history, law and social studies may also have political relevance. As such, the school socialization argument suggests that the impressions students form of politics within these learning environments may have lasting effects on future political behaviour. To what extent do these courses pique the interest of the students in politics? To what degree do these learning experiences shape expectations of future voting? While these effects may not be definitive, our expectation is that students who have positive feelings about political learning in school will be more likely to see themselves as future voters. The variable we use to represent respondents' attitudes toward in-school political instruction is based on their level of agreement or disagreement with the statement “On the whole, learning about politics in school has been enjoyable.”
It is important to note that school-based political socialization experiences extend beyond formal curricula, as well as beyond the confines of the classroom. While our focus in this article is on students' attitudes toward formal politics instruction in the classroom, the high school experience provides far more in the way of political socialization than is represented by this variable. Adolescents learn politically relevant lessons about power, hierarchy, and social class both in and outside the classroom, for example. These experiences comprise an important element of political socialization, broadly understood, a point made aptly by work on the so-called “hidden curriculum” (for example, Bhavnani, Reference Bhavnani1991; Gordon et al., Reference Gordon, Holland and Lahelma2000).
The third set of factors that we consider deals with political attitudes, specifically, the effect of adolescents' political interest and cynicism on the likelihood of seeing themselves as future voters. We expect higher levels of political interest to be associated with greater levels of anticipated voting, similar to adult populations across the industrialized West (Franklin, Reference Franklin2004). Cynicism has received a lot of attention in recent years, particularly in light of dramatic downswings in turnout. While findings suggest that youth are no more cynical than adults and may even be less so (Blais et al., Reference Blais, Nadeau, Gidengil and Nevitte2002; Gidengil et al., Reference Gidengil, Blais, Nevitte and Nadeau2004; Nevitte et al., Reference Nevitte, Blais, Gidengil and Nadeau2000; O'Neill, Reference O'Neill2001; Pammett and LeDuc, Reference Pammett and LeDuc2003; Rubenson et al., Reference Rubenson, Blais, Fournier, Gidengil and Nevitte2004), there is an important strain of cynicism pervading youth attitudes toward politics. We expect that the most cynical youths will be least likely to anticipate turning out to vote once they become eligible. Our models include two types of political disaffection: cynicism toward political parties and cynicism toward politicians. Cynical attitudes towards parties are measured by responses to the statement, “all parties are basically the same; there isn't really a choice”; for politicians, respondents agreed or disagreed with the statement, “politicians are ready to lie to get elected.”Footnote 13 For both variables, response options comprise a scale from “strongly agree” to “strongly disagree.” Given that cynicism toward politics and politicians is relatively widespread, it is likely that intentions to become a voter will be lower among the more cynical adolescents compared to the less cynical.
The last factor we incorporate in our analyses is political knowledge. The positive relationship between knowledge and turnout is one of the most robust and enduring in the political behaviour literature (for example, Delli Carpini and Keeter, Reference Delli Carpini and Keeter1996; Gidengil et al., Reference Gidengil, Blais, Nevitte and Nadeau2004; Popkin, Reference Popkin1991; Popkin and Dimock, Reference Popkin, Dimock, Soltan and Elkin1999). As such, we expect that the highly knowledgeable will be much more likely than the least knowledgeable to see themselves as future voters. Our political knowledge variable is an index; scores are assigned based on respondents' answers to 10 political knowledge questions.Footnote 14 Each additional knowledge question answered correctly increases a respondent's score by one point to a maximum of 10 points for a perfect score.
While a variety of additional explanatory variables have potential relevance for our central research question, exclusions were necessary based on the focus of the paper and the availability of appropriate indicators. For example, data on parents' and teachers' own attitudes and political engagement could add richness to the analysis through permitting us to tease out the effect of various agents of socialization. In line with recent work by Johnston and his colleagues (Reference Johnston, Matthews and Bittner2007), the inclusion of survey questions about respondents' perceptions of political competitiveness (or uncompetitiveness) would have likewise permitted us to test the theory that the nature of political competition during one's youth affects attitudes toward turnout.Footnote 15
We model the effects of these variables on future turnout in a sequential manner.Footnote 16 The future voter model is estimated first as a baseline model that includes only socio-demographic variables. Variables representing socialization, political attitudes and political knowledge are introduced in subsequent estimations. We proceed in this fashion based on the theoretical assumption that there is a temporal logic to factors affecting anticipated turnout. In this sense, we suggest that there is a causal ordering of effects on anticipated turnout and some variables are both further from the turnout decision (for example, gender) and work through variables introduced later in the models. By contrast, we contend that other variables (such as cynicism or political interest) are closer to the turnout decision and are a function of the prior variables included in the model (such as various agents of socialization).Footnote 17 Models are estimated using logistic regression because the dependent variable is dichotomous.
Results
Distribution of Dependent Variable
Starting with a distribution of the dependent variable, expected future turnout, a large portion of adolescents would vote if an election were held tomorrow and they were eligible. As Table 1 indicates, among those students who had clear intentions about future turnout, almost 78 per cent would vote, and about 22 per cent would not vote. This distribution is nearly identical to future turnout reported by 14-year old Americans (Hooghe and Stolle, Reference Hooghe and Stolle2004).
Table 1. If there was a provincial or federal election tomorrow and you were eligible to vote, would you vote?

Source: The Students' Assembly Survey on Youth Attitudes to Democracy and Political Participation
While one reading of this distribution could be optimism for ameliorating turnout decline among youth, caution is warranted. Over-reporting of turnout is perennial (for example, Karp and Brockington, Reference Karp and Brockington2005). Among respondents to the 2006 Canadian Election Study, about 90 per cent claimed they voted in the federal election, while actual turnout was just under 65 per cent.Footnote 18 Over-reporting of turnout is due to self-selection of survey respondents and social desirability bias. Predictably, people who vote are more likely to participate in political surveys due to interest in politics or feelings of duty to contribute to political life. Additionally, because voting is a socially desirable behaviour, there are always non-voters in a sample who do not want to reveal their abstention.
A second reason to be cautious about the high rates of future turnout concerns the context in which the survey was conducted. For the most part, students completed the survey after a unit of classroom instruction about electoral reform. It is possible that students were primed to think about political participation, and the unit on electoral reform may have stimulated political interest that would not otherwise exist in the sample. While this potential bias is important to bear in mind for our data analyses, in the larger scenario, if political instruction stimulates participatory attitudes among adolescents, civic education may help reverse youth turnout decline. Indeed, this has been part of the rationale behind the (re)introduction of high school civics classes in Ontario and other provinces in the last few years.
Future Voter Models
Results from regression analyses are presented in Table 2. As previously discussed, estimations proceeded in four stages whereby socio-demographic variables, socialization variables, attitudinal and cognitive variables were introduced sequentially as blocs. Starting with the impact of grade level, model 1 suggests that our prediction is correct. An increase of one grade level (from Grade 10 to Grade 11, for instance) is associated with a 40 per cent increase in the likelihood of future voting. This result may reflect the fact that politics may be more relevant to adolescents nearing voting age. In addition, the senior students had completed the mandatory Grade 10 civics course by the time our survey was administered, and this course may have boosted feelings of civic duty among these older students, which is an important predictor of turnout (Blais, Reference Blais2000). Results from model 1 also suggest that Catholic youth are 85 points more likely to vote than their non-Catholic counterparts. This suggests an enduring influence of Catholicism on turnout, even among the most secular segment of the population. Gender has no significant effect on future turnout, as predicted. Finally, the very small pseudo-R2 (0.03) indicates that the model explains little variance in the dependent variable. Clearly, there are other factors at work.
Table 2. Logistic Regression of Anticipated Turnout among Adolescents

Source: The Students' Assembly Survey on Youth Attitudes to Democracy and Political Participation
Note: Dependent variable is whether adolescents would vote in a hypothetical provincial or federal election if they were eligible. Cells contain binary logistic regression coefficients, standard errors in parentheses, and odds ratios in italics.
a p < .01;
b p < .05;
c p < .1.
Model 2 introduces variables associated with socialization experiences: family political discussion, involvement in peer groups and experiences with political instruction in the classroom. Each of the socialization measures has a significant and positive correlation with anticipated turnout. Predictably, habitual political discussion in the family has the strongest effect on whether students see themselves as future voters, which is in line with the notion that family is the earliest and most influential agent of political socialization (for example, Hyman, Reference Hyman1969).
The family is not the only important agent of political socialization. Students who are involved in at least one peer group are 75 per cent more likely to see themselves as future voters than those not involved in at least one peer group. This result indicates the potential importance of associational activities for fostering participation in political life, a finding consistent with work on social capital and the potential for ostensibly apolitical organizations to politicize (Putnam, Reference Putnam2000; Erickson and Nosanchuk, Reference Erickson and Nosanchuk1990).
A final component of socialization included in the model measures adolescents' experiences learning about politics in school. Presumably, all things being equal, an enjoyable experience with political learning in school should (hopefully) dispose one to becoming politically involved when reaching adulthood. For each one-unit increase (on the four-point scale) in learning enjoyment, the likelihood of seeing oneself as a future voter increases by about 58 per cent.Footnote 19 This result is interesting for a few reasons. First, it demonstrates the importance of school experiences on the development of political attitudes and behaviours. When students enjoy learning about politics, they become more inclined toward political participation. From a public policy perspective, this result suggests that addressing the quality and nature of high school political learning experiences may be a means for increasing future turnout among today's youth.
The third model incorporates a bloc of attitudinal variables. Popular discussion of turnout decline tends to lay the blame on disaffection and cynicism toward politics. Our results partially support this logic. For each one-unit increase in cynicism towards political parties, the likelihood of future turnout declines by 42 per cent.Footnote 20 By contrast, cynicism toward politicians has no independent effect on anticipated turnout among students. These contrasting results are interesting and deserve further discussion.
To start, general cynicism towards political parties is not greater than cynicism toward politicians. On a 0 to 3 scale where 3 is most cynical, the mean score for cynicism towards parties is 1.03, and the mean for cynicism towards politicians is 2.14. These adolescents are much more cynical toward politicians than they are of parties. However, what appears to drive declining intentions to vote is not cynicism writ large or cynicism toward politicians, but rather cynicism towards parties. As such, the youth who are most cynical toward parties are also the least likely see themselves as future voters. Why do cynical attitudes toward parties, but not politicians, play such an influential role in anticipated turnout among youth? Perhaps youth feel that none of the parties adequately addresses the issues most important to youth, giving youth a sense that parties are “all the same” in their relative neglect of young people.
Unsurprisingly, politically interested youth are more likely to see themselves at the ballot box in the future. Moving up one unit on the political interest scale (coded 1 to 10, representing ascending levels of interest), the likelihood of seeing oneself as a future voter increases by about 28 per cent. While socialization likely precedes political interest (particularly among youth), higher levels of political interest increase anticipated voting, controlling for all other variables in the model and independent of prior socialization effects.Footnote 21
Finally, the addition of political knowledge in the final model produces unexpected and perplexing results. Most work demonstrates that knowledge has a dramatic influence on political behaviours, and boosting turnout is one of the most consistent effects of higher political knowledge (for example, Delli Carpini and Keeter, Reference Delli Carpini and Keeter1996; Gidengil et al., Reference Gidengil, Blais, Nevitte and Nadeau2004; Milner, Reference Milner2002; Popkin, Reference Popkin1991; Popkin and Dimock, Reference Popkin, Dimock, Soltan and Elkin1999). Results from model 4 suggest, however, that knowledge has no significant effect on whether adolescents would vote in an election if they were eligible.
The unanticipated result for our knowledge index warrants further investigation. Looking at a cross-tabulation of political knowledge and anticipated turnout, there is an apparent positive relationship between the two. As Table 3 indicates, compared to just under 70 per cent of low knowledge students, 88 per cent of the high knowledge students see themselves as future voters.Footnote 22
Table 3. Cross-tabulation of Knowledge and Expected Turnout

Source: The Students' Assembly Survey on Youth Attitudes to Democracy and Political Participation
Note: Cells contain percentages with raw frequencies reported in parentheses.
Pearson's Chi Sq = 19.65
p < .001
Cramer's V = 0.17
Since the hypothesized relationship does show up in tabular analysis, what accounts for the non-performance of knowledge in the fully elaborated model? The first issue we address is the possibility of high multicollinearity in the model, and while this does not seem to be a major concern,Footnote 23 the fact remains that knowledge is correlated with other variables in the model, which is unavoidable in social and political research. Of the variables included in model 4, knowledge is correlated with family political discussion (0.42), school learning experiences (0.32) and political interest (0.48) (all significant at p < .001). These intercorrelations may be sufficiently substantial to obscure a significant relationship between knowledge and future turnout.
In addition, we have to bear in mind that our focus is on a specific sub-group of the population: adolescents in Grades 10 to 12. At this stage of political development, the correlates of political participation may simply be different than those among the general adult population. In this case, other factors may be more important for adolescents' attitudes toward voting than cognitive resources. During this period of political development, levels of respondent knowledge are largely a function of prior factors, in particular, socialization and political interest. As our model ordering suggests, we believe the effects of socialization are causally prior to both political interest and knowledge such that political interest is in part a function of socialization, and knowledge is a function of both socialization and political interest. Stated differently, whether adolescents see themselves as turning out to vote may be driven principally by socialization experiences. In turn, these experiences have prominent effects on political interest and finally knowledge. As a result, the hypothesized independent effects of knowledge fail to materialize when the model includes the range of factors linked to knowledge acquisition.
Discussion and Conclusions
In considering the topic of youth turnout decline more generally, this paper has sought to give heightened attention to the pre-adult years, “the period of maximum change” (Niemi and Hepburn, Reference Niemi and Hepburn1995: 7). We use this final section to situate our findings within the youth turnout literature and consider possible implications for future work.
One central finding suggests that expected turnout among our sample of adolescents is relatively high. In terms of adolescents' overall anticipated turnout levels, our findings are virtually the same as those reported by Hooghe and Stolle (Reference Hooghe and Stolle2004); nearly 80 per cent of adolescents who express a clear opinion about their anticipated adult political behaviour see themselves as future voters. By comparison, Elections Canada's estimates of young adults' (age 18–29) turnout in the 2004 Canadian federal election indicate actual turnout to be around 40 per cent, much lower than that anticipated by our adolescent sample.Footnote 24 This discrepancy, of course, raises the question of why adolescents' anticipated turnout rate is so much higher than the actual turnout of Canada's youngest electors in recent years. The discrepancy suggests that young adults want to vote and, in fact, intend to vote once they become eligible. Yet, when the time comes, young adults have abstained in surprisingly great numbers. One approach to definitively understanding this puzzle may be to conduct a panel study of Canadian youth whereby they are surveyed in high school and again during their early twenties (after one or more federal or provincial elections). This type of research design would provide greater leverage on understanding why youth turnout declines.
Our findings also raise questions about the correlates of turnout. Indeed, a core finding of the paper is the central role of socialization in shaping attitudes and expectations about future political involvement among youth. Family political discussion, peer group involvement and positive experiences learning about politics in the classroom all boost anticipated turnout among students. If, as findings suggest, socialization variables are key in the formation of political attitudes among youth, future analyses must seek to further understand the nature of these socialization influences. For instance, while the extent of family political discussion is clearly important, other features of the family context—such as whether the parents vote, are members of political parties or are members of other community groups—may also be salient in shaping youth political attitudes. Clearly, there is much work to be done in further sorting out the effects of socialization, including its dynamic nature.
While agents of socialization are important influences on the development of youth political attitudes, a salient point to emphasize is that socialization is a politically neutral concept. While socialization tends to be thought of in normative terms as boosting healthy attitudes toward democracy and participation, it is possible that young people are exposed to negative forms of socialization that encourage apathy and cynicism toward politics (Gimpel et al., Reference Gimpel, Schuknecht and Lay2003: 13). While our socialization variables might be more aptly described as “positive socialization experiences,” negative political socialization experiences are possible and likely discourage anticipated political behaviour among youth. Negative attitudes toward politics can be acquired in homes where politics is discussed frequently, from news that trivializes or criticizes the political system or portrays politics as little more than a horserace, or from peers who send messages that earnest political engagement is “uncool.” As a result, future analyses of socialization must seek to distinguish between the “positive” and “negative” nature of these influences.
The most curious result is the insignificance of political knowledge as a predictor of anticipated turnout. Previous work has clearly established political knowledge as one of the strongest and most consistent predictors of political participation (for example, Delli Carpini and Keeter, Reference Delli Carpini and Keeter1996; Gidengil et al., Reference Gidengil, Blais, Nevitte and Nadeau2004; Howe, Reference Howe2006; Milner, Reference Milner2002; Popkin, Reference Popkin1991; Popkin and Dimock, Reference Popkin, Dimock, Soltan and Elkin1999). For Delli Carpini and Keeter, “political information is to democratic politics what money is to economics; it is the currency of citizenship” (Reference Delli Carpini and Keeter1996: 8). Therefore, contradicting the conventional wisdom that people with more information are better equipped to make political decisions, results in this article imply that this may not be the case among adolescents. Since adolescents' political attitudes and knowledge are still at a formative stage, the independent effects of knowledge may be tied too closely to factors that boost knowledge (such as socialization). In fact, this may serve as a caution to researchers against assuming that theoretical and empirical models of adult political behaviour can be automatically applied to adolescents.
We believe the findings of this paper are indicative of the value of moving backwards from a focus on initial turnout decisions made by new political adults to consider more seriously the pre-adult roots of participatory attitudes and behaviours. If initial decisions to go to the polls become path dependent behaviours resistant to change later in life (for example, Blais, Reference Blais2000; Franklin, Reference Franklin2004; Gerber et al., Reference Gerber, Green and Shachar2003; Johnston et al., Reference Johnston, Matthews and Bittner2007; Plutzer, Reference Plutzer2002), the years immediately preceding the vote acquire new significance.
Acknowledgments
An earlier version was presented at the 2007 conference of the Canadian Political Science Association in Saskatoon, Saskatchewan. We gratefully acknowledge the valuable comments on earlier drafts provided by Keith Archer and André Blais. We wish to thank the organizers of the Ontario Students' Assembly on Electoral Reform who collected the data used in the article as well as Catherine Slakmon for the French abstract. Finally, we thank the Journal's anonymous reviewers for their valuable comments and suggestions.