Mobilisation through early activation and school engagement – the story from Scotland

ABSTRACT This paper explores the experience of 16–17 year olds participating in the Scottish independence referendum and discusses whether it can be seen as positive or negative considering civic attitudes and participation. Using data from two comprehensive and representative surveys of 14–17 year olds, it engages empirically with claims about young people's alleged political (dis-)interest and provides qualifications for commonly believed stories of young people as mere recipients of information given to them by parents and teachers. The paper develops a positive view of young people's engagement in the referendum process and suggests that inputs from parents and schools actually have distinguishable effects on young people, who do not simply ‘follow the lead’ of others uncritically. The analyses suggest that the discussion of political issues in the classroom (rather than the simple delivery of civics-style classes per se) may act as a positive factor in the political socialisation of young people, but suggests that further research is required to examine these effects beyond the specific context of the Scottish independence referendum in particular in relation to questions about whether reducing the voting age to 16 could be expected to generally lead to positive outcomes.

In most countries in Europe voting in national elections is reserved for people who are least 18 years old. A few exceptions exist however: in 2007 Austria changed its laws and now allows 16 year olds to take part in national elections. Apart from this however, other European countries, such as Estonia or parts of Germany, tend to limit extensions of the voting franchise to 16-year olds to local or regional elections or small-scale trials (such as Norway).
Because of this data about the political attitudes of 16 and 17 year olds is relatively rare. Many of the studies that have engaged with discussions about the political participation of 16-year olds have therefore relied on studying young adults (often 18-24 or 30 year olds). That however comes with a range of problems as it assumes that these young adults are similar to their slightly younger counterparts. However, the vast majority of 16 and 17 year olds are still in schooling (in traditional full time education or vocational arrangements)and several studies have shown that being in school could have strong distinguishing effects in relation to young people's experience of political discussions and engagement (e.g. McDevitt and Chaffee 2000; Dassonneville et al. 2012;Stewart et al. 2014).
Additionally, as studies from Austria have shown (Zeglovits and Aichholzer 2014) engaging with young people's political attitudes may reveal different results when they actually have a vote at this age range already. So the limited range of good quality survey data on political attitudes of 16-17 year olds is exacerbated by the lack of opportunities to study their views and behaviour, as well as the influences that shape them, in a context where those young people are asked to take part in the democratic process.
A new opportunity to conduct such research has arisen given the Scottish independent referendum held in September 2014 for which the voting was lowered to 16. As the plans for this change were announced in 2012 already and there had been a long campaign leading up to the referendum, we have been able to investigate the young age group's political attitudes in the real context of an imminent vote in which they could take part.
This paper engages with data from two cross-sectional representative surveys of under-18 year olds living in Scotland who were interviewed in the context of the referendum on independence. Using this data the paper explores how political attitudes in this age group are shaped in the real context of a vote and discusses how different influences (in particular family and schools) impact those attitudes. The analyses provide an optimistic picture about young people's engagement with politics at the age of 16, in particular when they take part in certain, deliberative forms of civic education. The research contributes to the literature emphasising the important role of schools in the political socialisation of young people and makes suggestions for how the findings could be used as a foundation for further research into the specific role of the earlier enfranchisement at age 16.

Approach and review of core arguments
The purpose of this paper is to (i) explore whether young people in Scotland expressed attitudes and behaviour more similar to the optimistic or pessimistic expectations voiced about them and (ii) which factors (in particular families and schools) showed substantial impact on their civic attitudes. Discussions about 16-17 year olds' political attitudes and participation are often connected to arguments for and against lowering the voting age. While proponents tend to paint a very positive image of the earlier participation of young people, suggesting that young people should be seen as ' … societal assets that should be supported to develop their greatest potential, rather than treated as latent problems or sheltered from interactions with their communities' (Flanagan and Christens 2011, 6), opponents suggest that earlier enfranchisement may (i) have negative effects or (ii) not do any harm, but also not improve anything in terms of civic participation of young people. Below, I present a brief review of core concerns raised in relation to young people addressing both domains and opposing views presented by other researchers.
Electoral participation of young people will have negative consequences Many commentators have put forward concerns about 16-and 17-year olds' ability and willingness to properly take part in elections (Russell 2014). They suggest that extending the franchise would lead to letting a group of people take part that ultimately is neither interested nor informed enough to make valuable decisions. This sort of argument tends to be supported by two main sets of evidence: Findings about lower voter turnout in young age groups and higher levels of political delineation from traditional political actors coupled with lower rates of interest as a sign of political apathy. Indeed, we find a substantial gap in voter participation between younger and older age groups (Wagner, Johann, and Kritzinger 2012, 377;Electoral Commission 2014), though to a varying extent in different countries (Fieldhouse, Tranmer, and Russel 2007). The differences can be very extensive: In the Scottish context this was remarkably apparent at the 2011 Scottish Parliament elections where only 30% of the 18-24 year olds took part, compared with over 80% of those aged 65 and above (Eichhorn 2014, 11). That younger age groups are less likely to take part in elections is factual and hard to refute. The question is whether it logically implies that earlier participation of young people would be negative, as this apparently disengaged group should first be politicised during their early years of adulthood (a process we can indeed observe to some extent see for example Chan and Clayton 2006, 543). Therefore, critics argue, it would most likely just result in (i) a low turnout and (ii) a replication of their parents' voting behaviour. Experiments from Norway, in which young people in some local communities were given the right to vote in local elections, while others were not, suggested that there was no discernible positive effect on political attitudes for those who could take part (Bergh 2013). We also know that young people are less likely to associate themselves with a particular political party and that they often display lower levels of identification with traditional actors in the political system (Chan and Clayton 2006, 545), resulting in lower levels of knowledge about the system and its actors.
However, we need to take into account two important considerations to deepen our analysis: First, in order to use these points to characterise young people pessimistically, suggesting that their participation would result in negative consequences, the assumption is made that low voter turnout and disenchantment with political parties are signs of political apathy. Extensive research projects have demonstrated however, that such an assumption cannot be upheld, as many scholars have argued . With a decline in party membership and lifelong party allegiances in many Western democracies (Scarrow 2002;Dalton 2014, 195), we also saw a rise in the rates of participation in others forms of political participation (Welzel, Deutsch, and Inglehart 2005, 128) or community service engagement (Syvertsen et al. 2011) through interest group work, citizen initiatives, ad-hoc demonstrations and many other forms of elite-challenging political participation (Kaase and Marsh 1979, 149). Equating disillusionment with political parties with political apathy is careless and does not do justice to the complex engagement with different strands of the political process young people participate in. We therefore need to carefully examine young people's explicit political attitudes and the correlates thereof rather than inferring their views and motivations only from behaviour within a traditional electoral system, if we want to explore their actual political views.
Second, the assumptions made above all assume a distinctive causality in the argument. They see young people's attitudes towards politics, their knowledge thereof and related factors as variables that will influence their voting participation if they were allowed to take part in the electoral process at age 16 (and in more general terms, if this age group was studied explicitly). This is indeed how the Electoral Commission largely treated research into young people's voting behaviour and concluded for the voting age not be lowered in a report following a year-long consultation (Electoral Commission 2003a, b). However, this may be fundamentally flawed: It is entirely plausible that earlier political engagement could result (through direct or indirect effects) in an increase in political interest and engagement. This leads us to the second strand of arguments (Wagner, Johann, and Kritzinger 2012;Zandonella 2013, 1085).
Young people's participation at age 16 would not change future civic engagement Even if concerns about political apathy may not turn out to be validated and thus no negative consequences would be expected from the political engagement of young people, some argue that it should still not be emphasised, because it is unlikely to improve anything. Indeed, a policy should not only be implemented when it causes no harm, but when it generates outcomes deemed as desirable, in this case an increase in political efficacy and engagement of young people.
The problem however is, that the existing evidence base is limited. Crucially, many studies treat 16-and 17-year olds as a linear extension from 18 to 24 year olds, assuming that their behaviour would be most similar to this age group. However, this may be missing a point (Hart and Atkins 2011). Studies focussing on what changes occur in political attitudes and behaviour in 18-24 year olds are meaningful in their own right, but they can tell us little about what to expect about 16-and 17-year olds, yet much commentary is based on such evidence (see for example Johnson and Marshall 2004).
It is plausible, however, to expect substantial differences in the engagement of younger people: Most 16-and 17-year olds in the UK, and in Scotland, are in some educational institution (93% in our survey). This means that, in a situation like the independence referendum, where young people could vote at 16, we were able to observe their political attitudes during their first official vote while most of them were still in school. There is no similar setting at any point later in life. Universities and workplaces act as political socialisation agents as well, but school is a place in which, theoretically, all (or at least a very large number of) students could be exposed to engagement with politics and elections that may result in discernible effects (Schwarzer and Zeglovits 2013). None of these could be sufficiently examined by studying 18-24 year olds, as this situation could simply not be replicated in that age group.
Knowing that earlier voting experiences are habit forming (Dinas 2012), it is therefore imperative that we investigate real situations in which 16-and 17-year olds are given the right to vote and examine their attitudes and behaviour in such contexts, being able to analyse whether schooling in particular had a decisive effect. It is not common to find such occurrences. The experimental example from Norway mentioned earlier (Bergh 2013), which had no positive effects of lowered voting ages, can only act in a limited fashion as a marker, as it did not actually create a situation in which young people were enfranchised legally, but where they effectively participated, knowingly, in an experiment which changes the framework substantially.
We do however, have extensive evidence from Austria, where the voting age was lowered to 16 for all elections in 2007. This constitutional change was accompanied by changes in school curricula (Zeglovits and Zandonella 2013) to enhance learning about and engagement with the political process. Initial results showed indeed a first-time boost in participation and related attitudes. But crucially, over time, a lasting and positive effect could be observed in Austria on political participation and political attitudes adding to traditional lifecycle effects (Zeglovits and Aichholzer 2014).
It is clear that both family background and schools play an important role in the political socialisation of young people. Parents can have a positive effect on voting participation for young people still living with them (Bhatti and Hansen 2012). There is a relationship not only in terms of electoral participation, but general civic participation of parents and their children (Zaff et al. 2010, 607). However, to assume that parental influence is the only socialising pathway is inaccurate. Young people can also have an impact on their parents' civic attitudes, an effect that is particularly pronounced for young people who have been exposed to civics education (McDevitt and Chaffee 2000).
Schools do not compete with parents for political socialisation influences, they act in a complimentary and distinctive way. Different studies have shown that young people who take part in certain forms of civics education are more likely to show pro-civic attitudes, for example a willingness to obey the law (Torney-Purta and Wilkenfeld 2009), a greater willingness to vote when civic knowledge is increased (Torney-Purta et al. 2001) and political participation more broadly (Quintelier 2010). However, it is clear that the type of civic education matters greatly. While formal civic education through classroom instruction has for example been seen to positively affect political attitudes and behaviour, open classroom discussions on political issues were seen to be more important to increase political trust (Dassonneville et al. 2012). So formal instruction about civic issues is useful, but not sufficient in generating pro-civic outcomes for which active discussion-based formats are required (Torney-Purta and Wilkenfeld 2009).
In the rest of this paper we will examine what has happened in the real context of the 2014 Scottish independence referendum to see whether (i) 16-and 17-year olds' political attitudes were reflective of a more positive or negative view of young people and (ii) whether there was indeed a distinguishable effect of schools on young people's attitudes or whether parents were more important as socialising influences.

Data used
The data for this analysis stems from two surveys conducted as part of the 'Future of the UK and Scotland' Programme and was developed by a team of researchers from the University of Edinburgh. For the survey, 14-17 year olds living in Scotland were contacted. When the survey was conducted first (April and May 2013) a substantial proportion of 14 year olds would have become 16 by the time of the referendum and therefore 14 year olds were the youngest group of people included in the survey. The survey was repeated in April and May 2014 following the same approach. While all questions were answered by all participants, analyses focussing on decisions around the actual voting process are only analysed for those who would have been 16 for the referendum in this paper.
The survey was carried out as a telephone survey using random digit dialling techniques. At the point of the survey Scotland still had nearly an 80% landline penetration rate which made telephone surveys viable as a tool for genuine probability sampling when face-to-face interviews are too resource-demanding. When a household with a 14-17 year old was identified, initially one parent of the young person was interviewed and asked for permission to interview their child, as we were dealing with minors, but also asked about their own referendum voting intention and their educational attainment (and the attainment of their partner, if applicable). We could use this information to have a variable that somewhat reflects differences in background between young people. Subsequent to interviewing the parent the young person was interviewed.
The survey was stratified by the eight regions distinguished in the Scottish Parliament elections representing equal proportions of the electorate. In addition, the distributions of sex and parental educational attainment were monitored. As people with lower socioeconomic status tend to be less likely to possess a landline and are less likely to participate in surveys, we expected an over-representation of respondents whose parents had higher educational attainment. We therefore weighted the results for expected parental educational attainment. For this purpose we used pooled data from the 2012 to 2013 Scottish Social Attitudes to estimate the respective educational attainment proportions for people aged 30 or above who had at least one child living in the household with them. The weighting altered results slightly, but not majorly. In 2013 proportions for key variables (such as referendum vote intention) changed by up to 2 percentage points when weights were applied. In 2014 the distribution was even closer to the expected values and changes after weighting did not exceed 1 percentage point for any variables ). All results presented in this paper are based on weighted data.
The majority of questions used in the survey was taken from the Scottish Social Attitudes Survey (SSA) to utilise the high quality of the question design process for the SSA, including cognitive testing for new questions and comprehensive piloting and it allows for the comparison of results to the general population. The initial questionnaire was piloted with 110 14-17 year olds in focus groups for two reasons: The SSA is a face-to-face survey, while this survey was carried out over the telephone. Some questions that may work well with showcards may not work adequately when read out over the phone. Furthermore, SSA questions have been tested for adults, but there may be differences in how this specific target age group would engage with some of the questions. Following the pilot approximately one third of questions was altered to take into account feedback received from the participants to ensure questions would work well with the age group and in a phone setting. The samples contained just over 1000 participants each.

Method and approach
The analysis begins by presenting an overview of some key variables relating to young persons' engagement with politics and the referendum, specifically looking at their levels of political interest, the extent to which they discussed the referendum with others and the likely voting turnout in comparison to the adults as well. After establishing this descriptive overview the paper engages in more depth to understand which factors affected young people's political attitudes.
For this purpose regression models were used to investigate in particular which institutions (family and school) influence particular aspects of young people's political attitudes. A set of ordinal regressions was computed to understand what differences exist between young people in relation to (i) their general political interest, (ii) their self-perceived understanding of politics and (iii) their voting likelihood. At first socio-demographic characteristics were used as independent variables (sex, age and the highest educational attainment of the parents 1 ). Subsequently variables were added that capture whether the respondent had discussed the referendum with their parents or in their class at school. This variable is key to examine whether there were distinct effects of engaging with the issue under consideration in schoolwhich would suggest that there might be something distinctive about political socialisation while a young person is still in school, thus providing evidence that 16 and 17 year olds had to be looked at distinctively in comparison to older young people. Finally, a variable was added reflecting whether the respondent had ever taken Modern Studies in school at all, and if they had, whether they had taken it as a mandatory or an optional course. Modern Studies is a course that not all, but an increasing number of schools offer in Scotland. It combines elements from several social sciences, such as Sociology, Politics and Media Studies. Using this variable we could identify whether such a class that we may characterise as 'civics' education had a discernible effect in itself, that could be distinct to having discussed the issue under consideration actively in class, regardless of the subject context. This importance in distinguishing between formal civics education and discussions in class has been demonstrated in multiple studies (e.g. Torney-Purta 2002; Torney-Purta and Wilkenfeld 2009; Dassonneville et al. 2012). As the paper aims to understand the experiences of 16 and 17 year olds in the context of the lead up to the 2014 referendum, it would be helpful to be able to partially distinguish those specific effects (such as discussing the referendum with parents and in class) from general differences in political interest (the first of the three dependent variables). For the second (self-perceived understanding of politics) and third (voting likelihood) set of models we therefore added a final model configuration where we including the general political interest variable as a further covariate. This allowed us to check whether the specific influences in the context of the referendum were merely expressions of political interest differences or actually robust effects.
The dependent variable for the first set of regressions, political interest, was measured by the question: 'How much interest do you generally have in what is going on in politics.' The model was formulated as a logistic regression. 2 To measure the self-perceived understanding of politics respondents were asked how strongly they agreed or disagreed with the statement 'Sometimes politics and government seem so complicated that young people find it difficult to understand what is going on.' 3 The version of this question used before the pilot was formulated to say ' … that I find it difficult to understand what is going on', however, this was perceived as very condescending and patronising by the pilot participants which is why the formulation was generalised. Based on its five-point scale an ordinal regression was applied with a complimentary log-log link function, as a greater number of responses was found for higher values (those agreeing with the statement that politics could be difficult to understand). Finally, to examine factors related to the likelihood of voting another set of ordinal regressions was computed. Again, a complimentary log-log link function was applied as higher values (greater turnout likelihood) were significantly more common. The question was: 'In autumn next/this year there will be a referendum to decide whether or not Scotland becomes an independent country. Assuming you will have the vote, on a one to five scale, where one is very unlikely and 5 is very likely, how likely would you be to vote?' 4

Results
We find people with a greater affinity to politics and people with lower affinities in all age groups of the population. Considering 14-17 year olds living in Scotland, assertions that young people had lower levels of interest than adults could not be supported (Table 1). The results from the young persons' survey 2014 and the Scottish Social Attitudes Survey (SSA) 2014 cannot be compared perfectly, as the SSA contained a middle option (some interest), while the young persons' one did not. While we do not know how many young people on either side would have chosen a mid-point option, generally the results suggest that levels of interest for young people were not dramatically different to those of adults. Nearly the same proportions were found in the extreme categories (just above or below 10% respectively), saying they had a great deal of interest or no interest at all. For the adults 'some' was the modal option with about equal proportions on either side. For the young people, more chose the positive option ('to some extent') than the negative one. This finding is in line with other research suggesting that young people do not display lowers levels of interest, but different conceptions of what they consider to be part of the body of politics (Quintelier 2007, 177).
A range of commentators had claimed that young people did not engage in discussions about political issues. Such suggestions could not be supported in relation to the referendum. Only 7% said in April and May 2014 that they had not discussed the referendum with anyone so far (Table 2). About two thirds had talked to their parents or their friends respectively and a similar proportion had discussed the referendum in class. All of these increased substantially from 2013 levels. Most respondents had discussed it with more than one of these groups.
However, some may claim that talking is one thing and taking part in a vote another. For the referendum on Scottish independence however, the youngest voters already indicated a great intention to vote in April and May 2014 (Table 3). Seventy-two percent of  those eligible to vote said in 2014 that they were very or rather likely to vote, a further increase from 2013 (66%). With this young people still seemed to be slightly less likely than adults on average to take part, however, the differences appear to be rather small (Table 4). While there is no exact comparison to adult data as the SSA uses an 11-point scale to measure likelihood to vote, we can look at it indicatively. Eighty seven percent of adults then gave a score of 6-10 on the positive side of voting likelihood. This is somewhat higher than the score for the young age group, however, the under-18 year old eligible voters also had an explicitly neutral middle option that was not provided on a scale and which may have prompted some adult respondents to pick a comparatively 'lower' option. The best evidence we have on the actual turnout levels by age comes from a survey conducted by the Electoral Commission. While sub-samples of specific age groups are limited in sample size and therefore prone to sampling error, it still gives us a decent general insight into the age-participation pattern in the referendum (Electoral Commission 2014, 64). With overall participation at just under 85%, the 16-17 year old estimate indeed remained lower at about 75% according to the survey. However, it was substantially higher than that of the 18-24 year olds which had the lowest turnout at only 54% (though bearing in mind the limited sample size and resulting error margins). This evidence is in line with previous findings discussed above showing that younger first time voters may show higher levels of participation (Bhatti and Hansen 2012;Wagner, Johann, and Kritzinger 2012;Zeglovits and Aichholzer 2014). This indeed supports the argument that we need to investigate 16-and 17-year olds in a different context than the slightly older, post-school age group and suggests that we cannot draw inferences from those aged 18 and above for younger adolescents.
While we have seen that the newly enfranchised voters could not be described as disinterested or unwilling to engage, some might argue, however, that this suggests that the engagement of 16-17 year olds may bring no harm, but it does not show yet that there are any potential benefits either of young people participating politically at this stage. It is therefore crucial to engage in more depth with the question whether there was something  distinct that could have influenced the political socialisation surrounding the first vote of these young people, characterising them as different to those voting for the first time at a later stage. Indeed, the vast majority of those under the age of 18 were in some form of school education. It is therefore crucial to examine whether school played a distinct role in this process. When assessing the influence on political interest, those young people who discussed the referendum actively in class indeed had significantly greater levels of political interest (Table 5). However, the same is true for those who had talked with their parents about the referendum. As this is cross-sectional data we of course cannot establish causality directly. It may be that we capture young people talking more about the referendum because they are more interested in the first place (though panel-based research has found that the effect of political participation on political interest is stronger than the other way around: Quintelier and Hooghe 2011). While we cannot discern between the two possible directions of this relationship statistically, at least for discussions in class it is clear however, that they will usually not be based on the initiative of individuals themselves, as the topics discussed in the classroom most commonly are, of course, based on the agenda set by the teacher. A positive role for school could be found here, however, it was not distinct from the effect we found for discussions with parents.
Beyond the discussions with parents, socio-demographic background variables appeared to only play a small role. Those aged 17 were slightly more likely to be interested than the younger ones (though the difference was rather small). There was no significant difference between male and female respondents and also parental education did not appear to relate strongly to political interest. The actual practice of engagement seemed to be what mattered: While parental education did not relate significantly to interest, actually talking to parents did. We did find that talking about the referendum was more common in households with higher education, but the educational status itself did not affect political interest further. Similarly, simply the fact of taking a class in 'civics' education did not necessarily relate to greater political interest, while talking about the referendum in class did. Those who took Modern Studies simply because they had to were not more likely to have a greater political interesta positive relationship was only found for those who chose to take it themselves (though, again, we cannot draw direct causal inferences, as there may have been self-selection).
This difference between merely taking a class and discussing the referendum actively was also visible in relation the self-perceived understanding of politics young people had (Table 6). Taking Modern Studies, regardless of whether it was by choice or not, was not associated with reduced perceptions of politics as being too complicated. However, those young people who had discussed the referendum in class indeed were significantly more likely to say that politics was not difficult to understand for young people. Here the effect of school was distinct from parental effects. Those who had talked to their parents about the referendum did not feel any more confident in evaluating politics as understandable. So engagement with a political issue in class could have positive effects on building an understanding of politics it seems in a way that parents could not. There was some influence based on parents however. Those young people who had at least one parent with a university degree were less likely to agree that politics was too complicated to understand in comparison to those who had no parent with any level of formal educational attainment. Interestingly however, the differences in understanding between those with parents without any formal education and the other levels of higher educational levels were not significant. The effect is therefore not that encompassing. It is interesting to note that male respondents expressed greater levels of understanding than female respondentsalthough there was no difference in levels of political interest. We find similar results for adults generally in the SSA with women more likely to say that politics was complicated than men. Seeing this manifested already in such a young age group may suggest that we are dealing with certain gender socialising effects here. Adding the general political interest variable does not alter the results substantially (except for the difference between 14 year olds and those aged 15 and eligible to vote in the referendum becoming statistically insignificant). General political interest had a strong effect, being associated with greater levels of self-perceived understanding and the effect sizes of the relationships described above were reduced slightly, however, they remained robust, suggesting that the experiences in the context of the lead up to the 2014 referendum were indeed more than expressions of general political interest. While discussions in school seemed to positively affect the understanding of politics, while talking to parents did not, the opposite held for voting likelihood (Table 7). Those young people who had discussed the referendum with their parents were significantly more likely to say that they would vote. No such effect could be observed for discussions in class. Education of parents mattered again only in the contrast between those with higher education degrees and those with no formal attainment at all. Having said all this, we should pause to consider one assumption made: so far we have been treating discussions with parents and in class as independent from one another. However, the two were related to each other (Table 8). Those who talked about the referendum with their parents were more likely to have also talked about it in class and vice versa. We do not know of course whether in discussions with parents, the topic was brought up by the young person or the parent. However, while there may be some room for initiative for students to suggest topics for class discussion, in most instances the topic for a class will be determined by the agenda set by teachers. We may therefore expect that for discussions in class the majority were not initiated by the young person. Consequently, it is plausible that some of the discussions with parents could be consequences of young people having engaged with the topic in class first and then continuing to talk about it with their parents. It is impossible to assess this perfectly with cross-sectional data. However, we can check whether there was an effect of talking about the referendum in class when not including discussions with parents in the regression (model 4 of Table 7). Indeed, when discussions with parents were excluded there was a significant positive relationship between discussions in class and voting likelihood. It was smaller than the effect of talking to parents, so there clearly was a distinct effect of parents on young people's likelihood to vote. However, considering the authority of the teacher to set classroom agendas, it is likely that there was a partial indirect effect of talking about the referendum in class, that may enhance the effect of talking to parents on voting likelihood for those young people who did both. Similarly to the models in Table 6, we again found that including a variable for general political interest did not alter the results substantially. While effect sizes were reduced somewhat, all statistically significant results in the full model (3) were robust.

Conclusions
The newly enfranchised voters in the Scottish independence referendum were not simply following the lead of their parents or appeared to be easily influenced in an inappropriate way. Parents fulfilled an important role in socialising young people to vote as a norm. Furthermore, they had an influence on young people's levels of political interest. However, that did not mean that the youngest voters in this referendum did the same as their parents. Just over half (58%) reported the same position on the referendum question as the parent interviewed. When over 40% held a different view to a parent, we could hardly speak of a group that was simply taking on perspectives from others. Indeed, parents were not influential in the same way in all domains. Those young people who discussed the referendum did not feel significantly more confident about their understanding of politics compared to those who had not talked to their parents. It did not appear that parental insights were considered as factual or educational in that sense. However, discussions in class took on precisely that role. Where young people discussed the referendum in the classroom, their political confidence increased and the results were robust and appeared to reflect experiences beyond general political interest. Schools could perform an important function that was distinct to the position of parents and other actors. Some of the effects (such as on political interest and understanding of politics) appear to be direct. Crucially, they require the active engagement of young people in the classroom however. A simple taking of a civics-oriented class in itself is not enough. Furthermore, in addition to the direct links found, there are also indications of indirect links: Young people, who discussed the referendum in school, were more likely to have discussed it with their parents, too. So even influences on voting turnout may exist, but they are strongly contextualised within the parent-young person relationship.
In the Scottish context we did not find evidence to suggest that young people were less politically interested than adults and neither could we substantiate claims that young people were disengaged in the referendum of uncritically mimicking their parents' decisions. Through the special role schools can play for this age group still, there is a chance to positively affect young people's political confidence and knowledge. Overall, the experience of newly enfranchised young people's engagement in the Scottish independence referendum has to be evaluated as positive. Negative concerns about the participation of 16-17 year olds could not be verified, but indications for positive effects, in particular when connected to supportive schooling were found.
However, further research is needed to substantiate these points further. While these analyses demonstrate (i) that conclusions about 16 and 17 year olds should not be deduced from observations in older age groups, (ii) that we need to take into account the potential effect of genuine franchise extensions and related activities and (iii) that active civic education plays an important role in enhancing positive civic effects, further insights can be gained by building on this research. It is crucial to observe young people in real contexts of elections in which 16 and 17 year olds are enfranchised. Having said this, the research does not suffice to draw major generalisations about the lowering of the voting age itself for two reasons: first, the Scottish independence referendum was a distinctive experience and second, but related, we cannot tell from this data whether there are lasting effects beyond this special occasion. Two avenues for further research emerge: First, we should study the behaviour of the newly enfranchised young voters in Scotland in future elections to see whether we can observe a continuation of positive effects. Second, within the UK there is a distinctive opportunity to conduct a meaningful extension of this research in the form a quasi 'natural' experiment in which young voters in Scotland would be compared to young voters in the rest of the UK for upcoming elections. It would allow us to examine whether the effects of schools and parents could be replicated in other contexts as well, but importantly, it would also allow us to see whether observed relationships can be replicated for Scottish young people (effectively a 'treatment group') in non-referendum contexts and in contrast to young people in the rest of the UK (effectively a 'control group').
Notes 1. For this the highest educational attainment of either parent is used. If for example one parent has a higher education degree and one only has lower secondary education, the highest educational attainment of either parent is coded as higher education degree.
2. The dependent variable distinguished those who said a great deal of interest, or were interested to some extent from those who said they had not very much interest or none at all. 3. The answer options were disagree strongly, disagree, neither agree nor disagree, agree, agree strongly. 4. The answer options were very unlikely, rather unlikely, neither likely nor unlikely, rather likely, very likely.

Disclosure statement
No potential conflict of interest was reported by the author.

Funding
This work has been supported by the Economic and Social Research Council under Grant ES/ K006460/1.