Marketing Brexit: An exploratory study of young voter engagement in relation to the EU referendum

This article presents a study of young voter engagement in relation to the EU referendum—a democratic vote on the United Kingdom’s membership of the European Union. Using the marketing concept of engagement, we examine how young voters engaged cognitively, emotionally, and behaviorally with the referendum and its associated campaigns. A mixed-method study combining multiple-phase questionnaires, longitudinal social network analysis of Twitter, and sentiment analysis provides a rich empirical description of young voter engagement. The findings reveal that young voter engagement is multi-faceted and varies enormously across our sample, particularly for behavioral engagement online. We subsequently question the analytical relevance of the engagement construct for political marketing before developing a typology to classify young voters according to the variance and extent of their specific engagement profile, which we define as “Prototypical Engagement Persona.” We conclude the article by presenting an agenda for future research on young voter engagement.


Introduction
Young people were the focus of widespread media attention before and after the United Kingdom voted to leave the European Union. Young people were often characterized as alienated, mistrustful, and disinterested with political systems (Barrett & Brunton-Smith, 2014;Dermody & Hanmer-Lloyd, 2004;Macnamara, Sakinofsky, & Beattie, 2012). Indeed, young individuals were less politically engaged than every other age group yet given their perceived preference for membership in the EU compared to older generations, their absence from the vote may have been the deciding factor. This raises an important question: is voter turnout the singular measure of engagement? This has led a variety of authors (Boonen, Meeusen, & Quintelier, 2014;Fieldhouse, Trammer, & Russel, 2007;Henn, Weinstein, & Forrest, 2005;Henn, Weinstein, & Wring, 2002;Wagner, Johann, & Kritzinger, 2012) to call for further research addressing young voter participation and associated confounding variables. The aim of this study is to (1) explore how young people engaged with the EU referendum event on 23 June 2016 and (2) to contribute more broadly to a general theory of political event engagement. This can help hypothesize how young voters differ and what can be done to activate their interest through the marketing of political events.
First, we review recent work on political marketing and identify the call for further investigation into how young voters become (dis)engaged. Second, we discuss the increasingly influential concept of engagement in the marketing-specific literature and how engagement can inform our empirical study. Third, we outline the research design of our study including analytical process, followed with the discussion of our findings around the two research questions in the fourth section. Finally, the article concludes by presenting an agenda for future research on political event engagement.

Political marketing
Political marketing represents a sophisticated area of study, "beyond the black arts of propaganda" (Harris & Lock, 2010, p. 297) and amalgamates theories, concepts, tools, and frameworks transferred from diverse disciplines such as commercial marketing and political science (Harris & Lock, 2010;Hughes & Dann, 2009;O'Cass, 2001;Speed, Butler, & Collins, 2015). Political marketing has been defined as "a set of activities, processes or political institutions used by political organisations, candidates and individuals to create, communicate, deliver and exchange promises of value with voter-consumers, political party stakeholders and society at large" (Hughes & Dann, 2009, p. 244). According to Simons (2016, p. 4) in "recent years there has been an increased level of voter volatility which has been matched by an interest in understanding electoral behaviour." Existing research tends to focus on two broad perspectives namely behavioral and managerial (O'Cass & Nataraajan, 2003). Behavioral political marketing includes aspects such as voter behavior, engagement, and participation, whereas managerial political marketing tends to focus on the development and management of political parties, political brands, targeted campaigns, elections, and the creation of political communications designed to appeal to desired segments (Hughes & Dann, 2009;Needham & Smith, 2015;O'Cass & Nataraajan, 2003;O'Cass & Voola, 2011;Simons, 2016). Despite progress made within political marketing, many gaps remain (Needham & Smith, 2015;Nielsen, 2016;O'Cass & Voola, 2011). There are explicit calls for more empirical research of voter-centric perspectives including how young voters engage and participate in the political process and how engagement develops through time (Gibson & Cantijoch, 2013;Macnamara et al., 2012).
As a discipline, political marketing can only progress if it continues to develop new concepts or reapply advanced theories and frameworks Smith & Speed, 2011;Speed et al., 2015). This study will address the limited number of theoretical frameworks in this research area to assist political stakeholders in the practical study of young voter engagement. The primary aim of the research is to consider how engagement data might be used to profile and segment young voters to improve targeting. This area of inquiry is increasingly pertinent due to the "growing democratic deficit identified in a number of contemporary societies" including the growing sense of mistrust, alienation, and disengagement of young people and politics (Macnamara et al., 2012, p. 624). In the following two sections, we describe the engagement construct as discussed in the politics and marketing literature before introducing a working definition of political engagement.

Young voter engagement and participation
Previous research suggests that "young people are the most disengaged of all the electoral segments in Britain" and often feel alienated with politics (Dermody, Hanmer-Lloyd, & Scullion, 2010, p. 422;Nickerson, 2006). Young people (18-25 years) are less likely to vote, get involved in conventional political activities such as joining and supporting political parties, and they hold negative attitudes toward the electoral process (Barrett & Brunton-Smith, 2014;Macnamara et al., 2012). In addition, young people are widely seen as disenchanted with politics, distrustful and cynical of political institutions, and have limited identification with political parties (Dermody & Hanmer-Lloyd, 2004;Dermody et al., 2010). Young individuals are described as being alienated, cynical, and disengaged with voting, yet also being interested and knowledgeable about political processes such as elections and voting (Dermody et al., 2010;Macnamara et al., 2012;Nickerson, 2006). This contradiction suggests young people can be distrustful and critical about politics despite having a strong sense of "personal efficacy," "where they believe they are knowledgeable about political issues and are interested in the election" (Dermody et al., 2010, p. 430). Improved personal efficacy may counteract cynicism and distrust of political voting, which can lead to a deeper sense of personal involvement and engagement (Dermody et al., 2010;De Vreese, 2005;Pinkleton & Austin, 2002), but it is also clear that this is not a simple linear relationship. Few existing studies consider whether the type of election (e.g., referendum or national elections) has an impact on young voter engagement and participation (Quinlan, Shepard, & Paterson, 2015). Most political research tends to focus on national rather than intermittent elections like referenda and fails to provide robust conceptualizations of what political engagement actually is.
The terms voter "engagement" and "behavior" are often used interchangeably without clear definitions (Dermody et al., 2010;De Vreese, 2005;Mann & Mayhew, 2015;Pinkleton & Austin, 2002). Similarly, "engagement" has also become synonymous with "participation" or voting in elections (Bromley, Curtice, & Seyd, 2001;Ekman & Amna, 2012;Ikeda, Kobayashi, & Hoshimoto, 2008;Macnamara et al., 2012;Nickerson, 2006) and this ambiguity broadens rather than focuses the debate. A few researchers attempted to develop insights into the conceptualization of engagementparticipation (Barrett & Brunton-Smith, 2014;Bromley et al., 2001;Ekman & Amna, 2012) such as dividing participation into conventional (e.g., voting, election campaigning, joining a political party) and non-conventional behavioral actions (e.g., signing petitions and participating in political demonstrations, and communicating with politicians; Barrett & Brunton-Smith, 2014). Similarly, engagement can be considered through other different "cognitive-emotive" manifestations, for instance monitoring political communications such as newspapers, television, radio, or social media, following political affairs, understanding political processes and institutions, having feelings about political matters, or discussing politics with other people (Barrett & Brunton-Smith, 2014). However, Barrett and Brunton-Smith (2014) fail to acknowledge the interchangeable nature of conventional and non-conventional forms and neglect the interchangeable nature of engagement factors by explaining that these forms appear to be cognitive-emotive characteristics; thus, existing studies do not explain how the conceptualizations of behavioral participation and cognitive-emotive engagement were developed (Barrett & Brunton-Smith, 2014;Bromley et al., 2001;Ekman & Amna, 2012). Therefore, future research should aim to provide further insight into this area of study.
Recent attempts to utilize digital platforms to mobilize young voters and strengthen engagement have had some success at the ballot box Macnamara et al., 2012;Mann & Mayhew, 2015). This seems logical given the fact that young people are avid users of digital platforms (Macnamara et al., 2012). Indeed, studies of political engagement seem more likely to provide holistic insight into young voters' lives if they consider engagement across multiple marketing campaigns and channels, such as social media where large tranches of political discussion now occur. However, most research in this area suggests voter engagement is linked to future habitual voting patterns (Barrett & Brunton-Smith, 2014;Gorecki, 2013;Nickerson, 2006) and we return to the same issue identified earlier that existing studies fail to conceptualize engagement beyond ambiguous definitions Macnamara et al., 2012;Mann & Mayhew, 2015). We suggest therefore the need to understand young voter engagement and participation, as this will provide pragmatic solutions to disengagement, and allow for the development of long-term strategies to mobilize the young electorate Macnamara et al., 2012;Mann & Mayhew, 2015;Nickerson, 2006;Thananithichot, 2012). Despite recent progress in this area, this complex debate still remains and differs from context to context (Barrett & Brunton-Smith, 2014;Dermody et al., 2010;Thananithichot, 2012). To develop a conceptual understanding of young voter engagement, we believe there is a need to step back and revisit the "engagement" literature more broadly within the marketing discipline.

Engagement in the marketing literature
The "engagement" concept is complex and debated in marketing and across other disciplines including organizational behavior (Bowden, 2009;Dessart, Veloutsou, & Morgan-Thomas, 2016) and management (Sashi, 2012), where there is an agreement that engagement goes beyond any simplistic behavioral element. In fact, researchers suggest that both affective and cognitive aspects are necessities for customer "engagement" (Bowden, 2009;Sashi, 2012). A similar debate exists in the marketing literature where much of the pioneering work examining "customer engagement" (Brodie, Hollebeek, Juric, & Ilic, 2011;Hollebeek, 2011;Van Doorn et al., 2010) has focused on defining the concept and distinguishing it from other similar concepts such as involvement and interaction. Central to all of these works is the claim that "engagement" is a multi-dimensional phenomenon and that cognitive and/or affective dimensions of engagement make this concept different from mainly behavioral concepts (Bowden, 2009;Mollen & Wilson, 2010). Despite the debate on engagement dimensionality, it is generally accepted among scholars that customer engagement includes cognitive, affective, and behavioral dimensions (Brodie et al., 2011;Dessart et al., 2016;Hollebeek, Glynn, & Brodie, 2014). Table 1 presents a summary of recent empirical research that conceptualizes dimensions and definitions of engagement in the marketing literature.
Cognitive engagement is "a set of enduring and active mental states experienced by the consumer" (Dessart et al., 2016, p. 408). Specifically cognitive engagement is illustrated by thought processing and elaboration toward the object of engagement and thus incorporates a person's intentional states, opinion, and knowledge. Affective engagement relates to words such as "bond," "care," and "emotion" when consumers speak about a brand (Dessart et al., 2016). Data for affective engagement can be gleaned by listening to the way people speak about an object and examining verbatim for the presence of cues in the language chosen, for example, the emotional valence of words. Behavioral engagement is defined as a "behavioural manifestation toward a brand or firm, beyond purchase, resulting from motivational drivers" (Van Doorn et al., 2010, p. 255). In a political context, this would mean behaviors beyond merely voting.
Engagement behaviors manifest in the political context in a number of different forms including both conventional forms, which involve electoral processes, and non-conventional forms, which occur outside electoral processes. Following work that has examined customer engagement behaviors online (Poorrezaei, 2016;Schivinski, Christodoulides, & Dabrowski, 2016), we suggest there are also at least three different forms of behavioral engagement observed through use of social media: (1) broadcasting-discussing explicitly political topics in public; (2) subscription-deliberately selecting and following political actors online to receive direct news updates; and (3) dialogue-explicitly sending public messages to political actors to support or persuade.
This study modifies Bowden, Conduit, Hollebeek, Luoma-aho, and Solem's (2017) conceptualization for engagement dimensions and proposes a working definition specifically for voter engagement as illustrated in Table 2.
This working definition and identified gap in the literature led to the following research questions that will govern the research design of this study: 1. How engaged were young people with the EU Referendum? 2. Can multi-dimensional engagement data be used to profile and segment young voters?
The following section will focus on the research philosophy and approach, participant selection, and questionnaire to conclude with the analytical process.

Research approach
In this study, the engagement object was the 2016 UK-EU referendum-a one-off political voting event. The empirical data collected focused on the emotions, thoughts, and behaviors of each An individual voter's degree of actual devotion of energy, effort, and time to a political event participant in relation to voting. The aim of the research was to provide an explanatory account of the inter-relation between the engagement dimensions for young voters. We used a concurrent mixed-method approach to investigate engagement of young people with the EU referendum and ascertain whether engagement data could be used to profile and segment young voters. Our approach initially focused on inductive reasoning before combining datasets to explain voter engagement.

Research methods
Our research approach was informed by the idea that the voter engagement construct is dynamic (it changes throughout time) and multi-dimensional (Barrett & Brunton-Smith, 2014;Dermody et al., 2010;Thananithichot, 2012). The study was thus designed to capture longitudinal data where possible and combine data collection methods that relate to each specific engagement dimension. We employed two key research methods: questionnaires (before and after the referendum) and a social network analysis (combining a network analysis of participants' Twitter account relations and the content of their discussions online). Questionnaires were designed to capture cognitive and emotional responses to the EU referendum immediately before and after the event (1 week before and 3 days after, respectively). Both questionnaires combined a mixture of open and closed questions to probe for reactions to specific topics (Banyard & Grayson, 2017) and to give participants scope to express their reasons for voting (or not) and their emotional reactions to the voting process. The social network analysis was designed to examine behavioral engagement beyond merely voting. This included (1) observation of the behavioral relations between young people and political actors online (i.e., do people actually interact with EU referendum campaigns, politicians, or parties) and (2) examination of the content of young peoples' dialogue online for the presence of behavioral engagement (i.e., do people actually talk about political issues online either informally in conversation or formally through dialogue with political actors). The overall structure of the method can be seen in Figure 1. The use of a mixed-method approach can strengthen the investigative process by capturing additional insight into the phenomena (Creswell, 2014;Fetters, Curry, & Creswell, 2013). Thus, social media (e.g., Twitter) was employed to complement the online questionnaire (prior and post referendum). The use of Twitter to get insights from consumers is becoming increasingly common in the current research environment (Nunan & Yenicioglu, 2013). The process of gaining permission for this form of data collection can be an issue for researchers (Nunan & Yenicioglu, 2013) because people can react with hostility when they become aware that they are the object of study for researchers (Hudson & Bruckman, 2004). This reaction reflects some of the ethical concerns raised by data collection from platforms such as Twitter, including privacy, informed consent, and the difficulty of applying ethical theory to this context (Conway, 2014;Kozinets, 2002). Although data were only gathered through Twitter, and thus represents only one specific form of social media with distinct functional characteristics, Twitter is the "leading platform among microblogging forms of social media that provides a way of broadcasting brief posts" (Fischer & Reuber, 2011, p. 3). Also, most data from this source are publicly accessible for data scraping in contrast to other platforms such as Facebook, Instagram, and Snapchat (Nulty, Theocharis, Popa, & Parnet, 2016). Furthermore, Twitter enables users to "share" personal thoughts, attitudes, and opinions and follow individuals, organizations, and campaigns (Bulearca & Bulearca, 2010;Fischer & Reuber, 2011).

Sampling framework
Participants were purposefully selected using a non-probability sampling technique (Saunders, Lewis & Thornhill 2015). Purposive sampling involves selecting respondents based on characteristics of the population to address the overall research objectives (Alston & Bowles, 2007). Young people (18-24 years) are often characterized as alienated, mistrustful, and disinterested with politics compared with every other age group, complex in nature yet represent an untapped market to political parties and politicians (Barrett & Brunton-Smith, 2014;Charles, 2009;Dermody & Hanmer-Lloyd, 2004;Dermody et al., 2010;Macnamara et al., 2012). Furthermore, this is supported by explicit calls for further research and new insights on young voter engagement (Boonen et al., 2014;Fieldhouse et al., 2007;Henn et al., 2005;Henn et al., 2002;Wagner et al., 2012). Thus, young voters represented the target sample for this study and were required to be: (1) aged between 18 and 24 years, (2) UK resident, (3) holders of an active Twitter account, and (4) willing to share data with the research team. Therefore, no additional demographic data such as social, regional, and educational background were collected. This could have been used to build a profile of the sample and future studies should address this. Participants were given a £10 Amazon voucher as remuneration for their participation after the final phase of data collection. Sample attrition was relatively high across data collection phases (>20%), some of this was due to participants not having an accessible Twitter account from which data could be captured.
The advert was distributed across social media channels and university email lists with encouragement for participants to share and help to advertise the study. The questionnaire required selfselection by participants; however, the research advert was non-descript in reference to politics or specific behavioral patterns of social media use. Since its creation in 2006, Twitter has been used extensively to understand behavior in the fields of marketing and politics (Conway, 2014). Twitter is distinctive from other social media platforms in terms of both characteristics and growth (Bulearca & Bulearca, 2010). The essence of events can be transmitted immediately across the network (Jansen et al., 2009), which makes it an attractive context to recruit participants. As we adopted a non-probability sampling technique, the socio-economic status of the participants could not be guaranteed, but given the exploratory nature of this study and the focus on the engagement construct itself (rather than the relation between participant background and engagement) it was not a threat to the legitimacy of the method. In addition, it is important to highlight that we make no claims of generalizability of the findings as our research questions aimed to understand how young people engage with a political event. Therefore, the representativeness of the sample should be considered in future quantitative research. This is set out in point three of directions for future research section.

Questionnaire design and sentiment analysis
The questionnaire was completed by participants online and in their own time using Qualtrics software for data collection. Participants were required to complete the first research stage before the day of the vote (EU referendum polling day-23 June 2016), whereas they completed the second questionnaire within 3 days after the results were released. A total of 101 participants took part in the first questionnaire, and 78 of which were involved in the further second questionnaire.
During the second phase of the questionnaire, participants were asked the question "how would you describe the result of the EU referendum" and given the opportunity to provide an answer in their own words. The results were then computationally analyzed for emotional valence using an academic sentiment analysis tool named Sentistrength (Thelwall & Buckley, 2013;Thelwall, Buckley, Paltoglou, Cai, & Kappas, 2010). This tool gives each word in the response a positive and negative score (depending on the expressed emotional strength) which are then combined to give an overall valence. This score can be plotted on a continuum from −4 to +4 to give a sense of the overall quantitative strength of the voters' emotional engagement. The scale begins at −1 (not negative) to −5 (extremely negative) and 1 (not positive) to 5 (extremely positive)-zero is not used until the numbers are combined. The reason for the dual score is that recent research suggests that people can and do experience mixed valences of emotion which often appear to be in conflict (Berrios, Totterdell, & Kellett, 2015). This is of especial interest in the case of political engagement to help reveal cases where people may feel emotionally conflicted about participation.

Social network relations and tweet content analysis
A longitudinal social network analysis was conducted by computationally scraping data from the social network service Twitter. The dataset included social relations of participants and their personal tweets. The aim was to reveal behavioral aspects of voter engagement external to the selfreported data in the questionnaires and avoid attitude-behavior gaps which may occur when speaking to participants. This was therefore an attempt to present an etic account of voter engagement through observation.
Accounts were mined using Twitter's API (Application Programming Interface) and exported into Microsoft Excel for subsequent analysis. Account details included tweet history (up to the previous 3,000 tweets) and account following (who the users themselves follow). All three of these datasets were generated to examine evidence of political discussion online, either indirectly with followers or through direct conversation with formal political accounts (official MP, MEP, political parties, or referendum campaign accounts). The corpus of formal political accounts was created by the research team by collating lists of official EU referendum campaigns, members of parliament, members of European parliament, and major political parties in the United Kingdom. The process of compiling this list was done by cross-checking a list of MPs, MEPs, and parties in office for official Twitter accounts. This process was aided by open-access watchdog websites that monitor tweeting of political actors online (e.g., www.mpsontwitter.co.uk, 2016).
The number of respondent Twitter accounts successfully mined was 71. Lookup tables were generated in Microsoft Excel for analyzing the prevalence of informal political discussions, direct dialogue, and ongoing engagement (following relations). The lookup tables of official formal political Twitter accounts included 73 MEPs with active Twitter accounts (4 MEPs without), 554 MPs with active Twitter accounts (88 MPs without), and 12 party political and official EU referendum campaign Twitter accounts. The number of tweets mined across all of the accounts was 156,650 with a mean average number of 2,206.3 tweets per respondent. The text of the tweets were examined for mentions of political accounts either indirectly (discussed generally in public) or direct mentions (using the @ symbol which denotes a direct notification of the user in question). Where tweets were found containing mentions, the research team manually examined the tweets to validate the computational analysis.

Results and discussion
This section sets out the results from the study and links the findings back to the existing literature. We first examine political engagement for young voters around the EU referendum and discuss our findings in light of the marketing engagement literature to answer the first research question. Next, we analyze the extent young voters fit into qualitatively distinct engagement segments and provide an account of prototypical engagement characteristics, thus answering the second research question.

Young voters' political engagement
Question 1 of this study investigated how young people are politically engaged pre and post the 2016 EU referendum. This section is divided into three parts, specifically: (1) a comparison of aggregate statistics pre-and post-referendum, (2) a sentiment analysis of respondent comments, and (3) a social network analysis of political broadcasting, subscription, and dialogue initiated by participants.
Comparison of aggregate statistics for pre-and post-EU referendum. The "cognitive" results displayed in Table 3 demonstrate the variability of voters' intentions when comparing their intentions preand post-referendum. One of the most striking figures is that 84% of our young voters aged 18-24 years were strongly inclined to vote before the EU referendum and their willingness to vote in a second referendum increased up to 96% after knowing the referendum results. This tendency is consistent with participants' opinion. In fact, the majority of young voters (79%) believed the United Kingdom should remain a member of the European Union and in the case of a second referendum, 82% said they would choose remain. In contrast with the strong voting intention for a second referendum, only 86% of participants would be certain to vote in a future general election as indicated in the second half of Table 3. This suggests engagement with future referendums (intermittent elections) seems stronger than periodic elections (General election). Although it is beyond the scope of this research to investigate the rationale for this discrepancy, this is consistent with the notion that young people are less likely to vote and be involved in conventional political activities. Perhaps the non-conventional and unique nature of referendums (intermittent elections), having a different objective from General Elections, is the reason why more young individuals would vote in a second EU referendum (Barrett & Brunton-Smith, 2014;Macnamara et al., 2012). This suggests that young voting habits are "malleable" and the prospect of shaping or influencing the electoral biographies of young voters as a means to strengthen engagement for future electionsvoting will appeal to political parties and campaign groups (Barrett & Brunton-Smith, 2014;Gorecki, 2013). When comparing political preferences preand post-referendum for an upcoming General Election, participants were fairly consistent, with two exceptions. First, pre-referendum, in the 2015 UK General Election, 33% of participants voted Conservative and 23% of participants voted Labor; however, more participants preferred Labor (33%) than Conservative (31%) in a hypothetical future General Election. This preference changed again post-referendum where more participants indicated support for the Conservatives (44%) compared with Labor (28%). While the change in opinion polls between elections is not surprising, the trend from our study seems consistent with national General Election polling at the time (www.ukpollingreport.co.uk). Since this study was conducted, the UK political landscape has changed again as a result of the 2017 snap-General Election. Second, the percentage of participants who would not vote in upcoming General Elections dropped from 16% (pre-referendum) to 5% (post-referendum). This is interesting as more young voters would vote in future UK General Elections after taking part in the EU Referendum, which suggests actual engagement and participation can impact on future engagement and possibly habitual voting (Barrett & Brunton-Smith, 2014;Gorecki, 2013;Nickerson, 2006).
When asked about the key issues that influenced voting choice before and after the EU referendum, participants' answers varied. As Figure 2 indicates, there are some notable differences in the key issues identified. Prior to the referendum, 56% of participants identified the "economy" as the key factor for voting. However, following the referendum, 67% of participants based their decision on the "economy" regardless of whether they voted remain or leave. The second key factor for voting was the "free movement of people"-14% prior and 10% post. Therefore, the economy was a major deciding factor in the decision-making process when participants cast their vote. The findings suggest that most young voters in our sample were knowledgeable about political issues (Dermody et al., 2010) and evidenced some thought elaboration and processing in relation to the event (De Vreese, 2005;Pinkleton & Austin, 2002). Figure 3(a) and (b) presents a comparative account of participants' perceptions to key issues before and after the referendum. A range of topics that featured centrally in official EU referendum marketing campaigns were put to the participants for their personal view on whether the United Kingdom would be better or worse off in the respective area if the United Kingdom voted to leave the EU. Generally, participants perceived that the United Kingdom would be worse off if its citizens voted to leave the EU in terms of: economic impact (pre = 77%, post = 73%), impact on jobs (pre = 68%, post = 66%), United Kingdom's global influence (pre = 61%, post = 58%), and for the participants personally (pre = 57%, post = 62%). By contrast, participants thought the United Kingdom would be better off in terms of the price of goods and services (pre = 76%, post = 81%). The perceived risk from terrorism remained similar (pre = 56%, post = 54%). Overall, the results remained largely consistent in most categories across both questionnaire phases, preand post-referendum. Sentiment analysis. Following the referendum, participants were asked to provide one sentence to describe the result (see Table 4). The majority of young "leave" voters (n = 14) revealed a positive emotional valence, optimistic responses such as "sensational," "hopeful for the prosperity of the UK economy," and "I'm excited by what the future holds for the UK." On the contrary, the majority of young "remain" voters (n = 70) expressed negative pessimistic views such as "people of the UK fucked up," "disappointing and sad. People let lies and scaremongering influence them," and "a victory for isolation, xenophobia and the irrationality of the mob." "Remain" voters generally provided a more negative valence swing than "leave" voters when responding to the question "how would you describe the result of the EU referendum." For  LEAVE voters 14 0.43 1.50 "Fantastic" (Peter, did not vote a ). "A heaving roar from the disillusioned, which has started the metropolitan political class into finally listening to them" (Jack, Conservatives). "Hopeful for the prosperity of the UK Economy" (Louise, Conservatives). "Uncertain, challenging and slightly devastating, but change is good and can work this out to benefit us if everyone comes together" (Shelley, Conservatives). "A positive opportunity for Great Britain" (Anne, UKIP). "Sensational" (Julie, did not vote). "A success" (Sarah, Conservatives). "I am excited by what the future holds for the UK" (Nick, Conservatives). "The result was fair and I feel it was the right result (Olivia, UKIP). REMAIN voters 62 −1.26 1.54 "A complete shit show" (Lisa, Liberal Democrats).
"The people of the U.K. fucked up" (Rose, Labor). "A victory of misinformation and fear mongering" (Matt, did not vote). "A victory for isolationism, xenophobia, and the irrationality of the mob" (John, Labor). "Disappointing and sad. People let lies and scaremongering influence them" (Emily, Greens). "Shocking-those who voted leave believed a bunch of idiots who lied to them. Half of them probably voted because of immigration but were too stupid to realise that it probably wouldn't make much difference" (Melanie, Conservatives). "Irritating, but democratically reached, although the handling of the result and planning for the future has been a clusterfuck" (Teresa, Labor). REMAIN voters: frustration over generational divide 8 -0.63 1.85 "I would describe the result as disappointing and detrimental for my generation, b the vote was a result of a lot of propaganda and ignorance and unfortunately will impact my generation the most" (Caroline, Conservatives). "Disappointing for young people, who mostly voted to remain. The general feeling amongst my peers is that the older generation made the decision to leave which will have more of an impact on young people, and this doesn't seem fair" (Thomas, Labor). "Everyone voted for the good of themselves, the elderly for out and the young for remain. With more elderly voters resulted in leaving the EU" (Jane, Labor). "Absolutely gutting: I feel the outcome is unfairly voted on by an older generation that won't see the true effect of the consequences; I am disappointed many young people didn't vote at all and I am most upset that some have used Brexit as a platform to validate racist views" (Helen, Conservatives). "Gutted, such a poor decision made by the majority of elder population which the younger generation clearly didn't want" (Eve, Labor). "Dissappointing, especially considering the results of age groups voting ie the older generation majority wanted to leave and deciding the future of the younger generation that wanted to stay" (Ian, Greens). "As a student, heartbreaking" (Malcolm, Greens). "Not a good result for the youth of the country" (Abraham, did not vote).
UKIP: United Kingdom Independence Party. a Previous vote in the general election 2015. b Emphasis added.
"remain" voters, the mean average valence when scores were combined was −1.26 (SD, 1.54). For "leave" voters, the mean average valence when scores were combined was +0.43 (SD, 1.50). It is hardly surprising that "remain" voters were disappointed with the result and "leave" voters were pleased with the outcome of the referendum. However, there was also an element of uncertainty with some "leave" voters. Two participants regretted their leave vote when asked in the second questionnaire and indicated that they would have voted remain if given another opportunity. Many of the "remain" voters also questioned the legitimacy of the vote itself despite participating. This supports the idea that young voters can be critical and cynical of politics yet engage with the political process (Dermody et al., 2010;Macnamara et al., 2012;Nickerson, 2006). In addition, a broader theme emerged in which eight "remain" voters were frustrated or angry with the outcome but directed their frustration primarily at older generations. In total, 57% of voters aged 55-64 years and 60% of voters aged 65+ voted to "leave" the European Union (Clarke, Goodwin, & Whiteley, 2017). The negative emotional valence expressed by participants toward older people is exemplified by comments such as "gutted, such a poor decision made by the majority of elder population which the younger generation clearly didn't want" and "disappointing, especially considering the results of age groups voting i.e. the older generation majority wanted to leave and deciding the future of the younger generation that wanted to stay." Therefore, many young "remain" voters attributed the result of the EU referendum to the actions of the older generation. The consequence is a tense generational divide. This reinforces the complexity emotivecognitive characteristics of young voter engagement and participation.

Social network analysis: respondent broadcasting, subscription and dialogue
Respondent broadcasting. The results show that young voters expressed opinions on a wide variety of political issues online. Table 5 reveals that participants discussed political parties (61.97%) and political topics (83.09%) generally, but also were very active in discussing the EU referendum specifically (45.07%) and its associated marketing campaigns. "Labor" and "UKIP" (UK Independence Party) were the most discussed parties followed by the "Conservatives" and the "Scottish National Party." In addition, participants discussed key political topics such as "jobs," the "economy," "Brexit," "NHS," and the "350 million" that would be given to the NHS if the United Kingdom voted to leave the EU. This suggests young voters were actively concerned with politics in the lead up to the referendum and had also well-elaborated thoughts in relation to the event. This supports the idea that digital platforms can strengthen engagement and develop political interest Macnamara et al., 2012;Mann & Mayhew, 2015).
Twitter subscription. Each respondent pursued and subscribed to information about other people and organizations on Twitter. The analysis of social network relations and tweet content comprised 20,902 accounts (nodes), 24,422 relations (edges), and highlighted a large variety of overlapping relations between participants. This is to be expected for Twitter-which is typically used to connect with existing relations such as friends, family, and colleagues alongside other relations including celebrities, interest groups, activists, politicians, and parties. The mean average follows for participants (343.97) and standard deviation (SD = 270.50) demonstrate that despite large variance within the sample, it offers a wealth of insights into the interests and conscious choices of young people online.
The results illustrate that young people were not just discussing political issues or topics but were actively following political stakeholders such as elected representatives and political organizations. This is consistent with the notion that young people are knowledgeable of political issues and actively discuss politics (Dermody et al., 2010;De Vreese, 2005;Pinkleton & Austin, 2002). In total, 28.17% of participants actively followed at least one MP, 5.63% followed at least one MEP, and 12.68% followed at least one party or official EU referendum account. Social media is now a key tool in fostering relationships between young voters and political parties (Dermody & Hanmer-Lloyd, 2004;Dermody et al., 2010) and increases the likelihood of young voters being exposed to political marketing campaigns. It is clear that a large portion of respondents in this sample use Twitter as a source of political news to stay informed.
Dialogue. The findings reveal that there is formal dialogue between young voters and political stakeholders such as politicians and political interest groups. For example, 36.62% of participants communicated directly with MPs by using the @ sign on Twitter, 16.9% communicated directly with MEPs, and 26.76% communicated directly with party and official EU referendum accounts. Many participants communicated directly with Members of Parliament such as "remain" supporters "Tim Farron" (Liberal Democrat), "Karl Turner" (Labor), and "Anna Soubry" (Conservative). However, fewer participants communicated directly with "leave" supporting Members of Parliament nevertheless, "Daniel Hannan" (Conservative) and "Nigel Farage" (UKIP) were two exceptions. The results highlight the huge variance of behavioral engagement for young people in the context of political discussion and self-selected political networks. This reaffirms the idea that young people form a sense of personal involvement with the political process. The results also demonstrate that young people exhibit a range of non-conventional behaviors which illustrate deep political engagement (Barrett & Brunton-Smith, 2014;Dermody et al., 2010;De Vreese, 2005;Pinkleton & Austin, 2002).

Explanatory account of young voters' political engagement
The second research question aimed to understand whether multi-dimensional engagement data can be used to profile and segment young voters. The traditional marketing activities of audience segmentation, targeting, and positioning rely on the initial possibility of profiling individuals and classifying them according to distinct segments. Audience segmentation provides the basis for a greater degree of personalization in marketing communications. Young voters can be classified according to their behavioral, cognitive, and emotional engagement. But the voter engagement construct is a compound from three inter-related dimensions, and neither of these dimensions are necessary or sufficient to indicate engagement generally. With this in mind, we put forward "Overlapping Engagement Characteristics and Prototypical Engagement Persona Typology" (Figure 4). Each Prototypical Engagement Persona (PEP) is defined through the presence/absence of qualitatively distinct engagement characteristics (behavioral, cognitive, and emotional) which depict an individual's momentary state of engagement.
Findings show that engagement is processual so any given individual will inevitably pass through multiple PEPs within their lifetime, but the identification of these momentary snapshots of engagement helps to understand how young peoples' engagement with politics transforms throughout time. We discuss below PEP of the proposed typology and suggest further research avenues that could utilize PEPs to inform political marketing campaigns.
Responsive (cognitively, emotionally, and behaviorally engaged). A responsive voter is defined as being fully engaged cognitively, emotionally, and behaviorally with the political event, thus all engagement dimensions are important to the marketing of the event. These people are aware of the political event, have registered to vote, participate in the voting process, have clearly elaborated thoughts on the event (knowing their representative MPs, MEPs, and parties), and have an emotional response to the process. They not only possess an intentional state of voting in the future but also demonstrate behavioral engagement beyond the voting booth in the present-they discuss political issues, subscribe to news about political actors, and even engage directly with political actors to persuade others according to their own respective political agenda.

Latent (cognitively and emotionally engaged, but lacking behavioral engagement).
A latent voter is defined as being cognitively and emotionally engaged, but not behaviorally. These people have responded to the marketing of the political event, but fail to actualize their engagement through participation in the voting process or beyond the voting booth. These people are typically aware of the event, have clearly elaborated thoughts on the event, and have an emotional response to the associated marketing campaigns. They may possess a positive intentional state for future political participation, but this fails to translate into action in the present moment. Indeed, these participants are most likely to be the young people who say they "forgot" to register to vote when asked for the reason for non-participation.

Cynical (cognitively engaged, but lacking emotional and behavioral engagement).
A cynical voter is defined as someone cognitively engaged who fails to participate in behavioral engagement and lacks a discernible emotional response to any of the possible political outcomes. These people are aware of the event and have views about its consequences, but fail to act in response to it. The cynicism used to describe this category is not necessarily restricted to a cynicism of the event itself, as it may relate to a personal assessment of the young person's own moral status, that is, they may ask the question "is it right that I am participating in this event?" and subsequently lack the moral impetus or even confidence to participate.

Disaffected (emotionally engaged but cognitively and behaviorally disengaged).
A disaffected voter is defined as a person engaged emotionally, but not cognitively or behaviorally. Although the person is aware of the political event and may even feel strongly that the result is positive or negative, this is not evidenced in the reasons they give or observed through behavioral participation. The disaffected voter represents a challenge for policy makers and anyone trying to encourage widespread democratic participation. Instinctive (behaviorally and emotionally engaged, but lacking cognitive engagement). An instinctive voter is defined as being behaviorally and emotionally engaged but lacking a clear and reasoned position. These voters may be conflicted about participation, but feel that the event is important even if they have a limited ability to articulate a reason for participation beyond gut feeling or affect. The anticipation of negative consequences followed by an expression of confusion and disappointment after voting for a successful outcome illustrates an inability to clearly articulate reasons in an ordered manner. Although participation is seen as important to this category of young voters, they may be the most unresponsive to subsequent political marketing, especially if marketers are unable to identify the political priorities of voters because their primary goal is political union rather than coping with the presence of dissent in their social lives.
Reluctant (behaviorally and cognitively engaged, but lacking emotional engagement). The reluctant voter is defined as being behaviorally and cognitively engaged, but lacking an emotional commitment to any outcome in the political event. Reluctant voters refuse to express a fully emotional engagement with the process because they see the process as misinformed, illegitimate, or misleading from all directions (not merely because of partisan bias). Nevertheless, they participate in the process, often because they see democratic participation as important regardless of their perspective of the event.
The reluctant voter is aware of and knowledgeable of the political event (in this case also knowing local MP, MEPs, and formal campaigns), registers to vote in the event, votes, and discusses informal topics through broadcasting, subscription, and dialogue. However, there is a consistent reluctance to fully accept the political event and consequently they suffer emotional dissonance during participation. Whether emotional disengagement can act as a mechanism to cause behavioral and cognitive disengagement remains a potentially insightful avenue for future research with profound consequences, especially as it relates to protest votes and "spoiling the ballot." Floating (behaviorally engaged but not cognitively or emotionally engaged). The floating engagement persona is defined as a young person that exhibits some behavioral engagement, but lacks cognitive and emotional engagement. The floating voter is non-committal, expressing neither elaborated reasons nor emotional valence in relation to vote outcomes, but nonetheless exhibits observable evidence of behavioral engagement. The behavioral engagement in this persona could be present in the form of voting without expressing cognitive or emotional engagement or more likely, the behavioral engagement is seen in partial glimpses outside of the voting booth, such as registering to vote or discussing political issues in relation to the event.

Conclusion and directions for future research
At the heart of this article is the proposition that young voter engagement is multifaceted and complex. Our study reveals young people were engaged with political discussion, actively follow political accounts online, and directly communicate with political representatives and organizations. Furthermore, our study suggests current engagement remains connected with future engagement and highlights the challenges of sustaining long-term political interest due to the "malleable" nature of young voters' electoral biographies (Barrett & Brunton-Smith, 2014;Gorecki, 2013;Nickerson, 2006). This article concurs with the notion that young voters can be interested in politics yet remain critical and distrusting of political institutions, the establishment, and the electoral process. Our findings support the idea that young voters who are interested and knowledgeable of political issues can offset negative emotional valence that would otherwise prevent behavioral engagement (Dermody et al., 2010;De Vreese, 2005;Pinkleton & Austin, 2002). Nevertheless, this could be context specific as the study was framed around the 2016 EU referendum, which raises questions about future research and applicability beyond this case.
We suggest that young voter engagement is context and concept specific, in terms of how voters inter-subjectively conceptualize the event itself. Our findings contribute to the evaluation of voter engagement as a dynamic process which changes through time as voters change their level-degree of engagement from election to election. Engagement is a multidimensional construct with overlapping characteristics, with no single characteristic considered superior. One of the main contributions of this article is to use the engagement construct for the practical purpose of classification, which can subsequently be used for audience segmentation and targeting. The classification typology that we name Prototypical Engagement Persona (PEPs) is polythetic, but nonetheless helps to conceptualize the actual extent and variance of the three dimensions of engagement. The classification schema thus serves as a practical tool to help understand the complexity of young voter engagement and inform marketing planning. Below we list a series of research objectives which future research should attempt to focus on to develop the legitimacy and validity of the PEP typology: 1. Future research could adopt the PEP typology as grounding for a qualitative in-depth exploratory study to understand the apathy of young voters and their engagement with the electoral process following the 2016 UK-EU referendum. More specifically, the application of the PEP typology will support the examination the engagement-disengagement of young voters following the 2017 UK General Election and assess future voting intention and re-engagement of young citizens 18-24 years. This in turn could assess the applicability of the PEP typology as a mechanism to understand young voter engagement, disengagement, and re-engagement which remains topical and highly relevant area of future research. 2. Alternatively, future research could deploy the PEP typology as a quantitative scale to measure repeated citizen engagement dimensions across large samples, jurisdictions, or different demographic groups. Engagement should be understood as a temporal phenomenon and is easiest to conceptualize through repetitive political events such as general or local elections. We believe the efficacy of PEP profiling and segmentation, according to the three engagement dimensions, will offer greater efficacy and insight when repeated longitudinally with a sample group. 3. Demographic and psychographic data are routinely employed to segment and target political populations, but multi-dimensional engagement data have thus far not been widely used in practice. In addition, future studies should record a wide range of demographic data to develop a detailed overview of the sample. We suggest the efficacy of future approaches which utilize the engagement construct will hinge on understanding the relative proportionality of PEPs in a target population. This will require larger datasets than presented here to help understand the generalizability of the prototypical persona outlined. 4. Large-scale profiling and segmentation using engagement data raises the ethical issue of unwanted or potentially invasive surveillance. We suggest that future young voter engagement research should ensure informed consent and respondent feedback in any study. Such a standard is not only fundamental to preserving ethical conduct but also reveals the literacy and willingness of voters to be understood on such terms.

Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.