Young trendsetters: How young voters fuel electoral volatility

Sociological theory posits that social change occurs first and foremost among young people, who set trends that may eventually carry over to older citizens. This study examined to what extent this proposition applies to electoral shifts by comparing parties’ electoral gains among young (age <25) and older voters (age >24) in 21 Western established democracies between 1948 and 2019. An analysis of 219 national election surveys revealed that winning parties typically gained disproportionately among young voters. This youth bonus was even stronger for new parties, whose electoral breakthroughs were importantly facilitated by youths. Electoral shifts among young voters furthermore predicted similar changes among older citizens in the subsequent election. This indicates that young people are not only more sensitive to electoral trends, but that they can also set trends that eventually carry over to older citizens. Young voters should therefore be seen as important drivers of electoral


Introduction
The German sociologist Karl Mannheim (1928) theorized that young people play a key role as drivers of change in society. In his footsteps, many scholars have argued that social change occurs first and foremost among young citizens (e.g., Münchmeier, 1982;Prakke, 1959;Rogers, 1995). Indeed, youths are often the first to embrace the latest trends such as new music (Smith, 1994), new fashion (Behling, 1985), or new media (Newman et al., 2020). There are compelling reasons to expect that young voters could play a similar role in electoral shifts. At a young age, most voters have not yet developed strong attachments to political parties (Dinas, 2014). Unconstrained by such party loyalties, the youngest age group may be most likely to follow electoral trends by supporting parties that have momentum in any particular election. Reversely, young voters may be the first to turn their back on parties that have fallen out of grace. Youths may also facilitate the breakthrough of new parties due to this lack of loyalty to existing parties. In addition to being more sensitive to electoral trends, it is moreover conceivable that young voters could function as trendsetters by creating momentum for parties that eventually carries over to older citizens.
To what extent youths play such a driving role in electoral shifts has never been systematically examined. Previous research on young voters has typically focused either on their role in long-term electoral change through generational replacement (e.g., Van der Brug and Franklin, 2018) or on their formation of political preferences (e.g., Dinas, 2014). Consequently, little is known about how the unstable party preferences of individual young voters (i.e., individual-level volatility) manifest themselves in electoral shifts between political parties (i.e., aggregate-level volatility). The present study therefore compared parties' electoral gains among young (age <25; N = 47,993) and older voters (age >24; N = 509,139) in 219 national elections that took place between 1948 and 2019 across 21 countries in North America, Western Europe, and Australia. Specifically, this study examined (1) if parties that win vote share gain disproportionately (i.e., more than among older voters) among young voters, (2) if this youth bonus is even stronger for new parties, and (3) if electoral shifts among young voters predict similar changes among older citizens in the subsequent election.
These questions are crucial to understand the electoral behavior of young voters. When new or winning parties are popular among youths, scholars often hypothesize that this popularity could be driven by generational differences in attitudes (e.g., Norris and Inglehart, 2019) and similar interpretations are common in the media (e.g., Financial Times, 2017;The Huffington Post, 2016;Vice, 2020). The possibility that new and winning parties attract a young electorate precisely because they are new or winning is however rarely considered. Before we can attribute the vote choice of youths to generational attitude differences, we first need to determine if their electoral behavior may instead reflect a heightened sensitivity to trends. Examining how young voters fuel electoral shifts may furthermore contribute to our general understanding of electoral volatility and why it is increasing in many countries (Chiaramonte and Emanuele, 2017).

Young voters and winning parties
The first objective of this study was to examine if winning parties gain disproportionately among young voters. Such a youth bonus for winning parties should be expected based on four theoretical mechanisms. The first mechanism is the idea of life-cycle differences in party loyalty. In 'The American Voter, ' Campbell et al. (1960) reasoned that voters develop increasingly strong levels of party identification as they grow older. This idea was further elaborated by Converse (1969), who argued in his essay 'On Time and Partisan Stability' that voters' party loyalty increases as they age and spend more time participating in the electoral process. The Michigan scholars theorized that party identification creates a 'perceptual screen' through which citizens interpret new information. As voters age, they are exposed to an ever-increasing amount of information about political matters. Due to motivated reasoning, people have a tendency to interpret this new information in a way that confirms their political identity (Bartels, 2002;Kahan, 2016). Voters' party identification may therefore strengthen over time, which in turn increases the partisan bias in their processing of new information. Sixty years of research on this phenomenon has indeed shown unambiguously that older voters are more likely to identify with a political party (Campbell et al., 1960;Van der Brug and Franklin, 2018), more likely to be strong identifiers (Converse, 1969(Converse, /1976Markus, 1983), less likely to change their identification (Alwin and Krosnick, 1991;Hobbs, 2019;Jennings and Markus, 1984), more likely to hold identity-consistent beliefs (Converse, 1964;Stoker and Jennings, 2008), and more likely to vote for a party that they identify with (Van der Brug and Rekker, 2021). Another explanation for these life-cycle differences in party loyalty lies in the psychological effect of voting itself. Due to cognitive dissonance, people tend to rationalize their choice after casting a vote by strengthening their preference for the party they supported (Dinas, 2014). As a result, citizens may gradually start to view their party preference as a core component their social identity as the number of instances that they have voted for the same party increases. Several studies have shown that citizens who have repeatedly voted for a party are indeed more likely to identify with that party and less likely to switch their vote (Dinas, 2014;Jennings and Markus, 1984;Meredith, 2009;Gomez, 2013).
A second reason to expect that winning parties gain disproportionately among young voters lies in generational differences in party loyalty. Although the empirical literature suggests that the stronger party loyalty of older voters should be attributed mainly to life-cycle effects (Converse, 1976;Dassonneville, 2013Dassonneville, /2016Shively, 1979), there is also evidence that more recent birth cohorts are less likely to identify with a political party than older generations (Claggett, 1981;Dassonneville, 2016;Van der Brug and Franklin, 2018). A first explanation for such generational differences can be found in the decline of 'cleavage politics' and 'frozen party systems' (Lipset and Rokkan, 1967) that took place in most countries since about the late 1960s (Franklin, 1992). Citizens who were socialized before this transformation may have inherited strong and stable partisan attachments from their parents, whereas people who grew up thereafter may have formed more changeable party preferences after entering the electorate in an era of volatility (Van der Brug and Franklin, 2018). Another explanation for generational differences in party loyalty lies in the increased educational attainment of younger generations. Dalton (1984;Dalton et al., 2000) for example theorized that cognitively mobilized voters are less dependent on partisan cues to determine their vote choice and more capable of re-evaluating their party preference in every new election. Research on this thesis has shown that the least educated citizens, who often belong to older generations, are indeed least likely to switch their vote (Dassonneville, 2013;Van der Meer et al., 2015).
A third rationale for vote switching among young citizens involves life-cycle differences in the determinants of vote choice. A common explanation for the increased electoral volatility in many countries is that citizens have become less likely to base their vote choice on stable factors, while changeable determinants have become more important (Dalton et al., 2000;Franklin, 1992). This idea can be related to young voters because research has shown that the youngest citizens are least likely to base their electoral choice on long-term factors such as social-structural characteristics or ideology (e.g., Boonen et al., 2014). For example, an age-period-cohort analysis of Dutch election surveys revealed that citizens under age 22 are least likely to base their party preference on their social class, educational level, left-right identification, or core issue attitudes (Van der Brug and Rekker, 2021). This life-cycle effect can be explained by the fact that it takes most people until about age 25 to fully develop their social-structural characteristics (e.g., by finishing education and getting a job), their issue attitudes, and their ideological identity (Rekker et al., 2019). A study on Dutch youths for example revealed that the amount of variance in left-right identification that is explained by social-structural characteristics and issue attitudes surges from close to zero at age 13 to respectively 10% and 19% at age 23 (Rekker et al., 2019). Because young citizens have yet to develop the long-term determinants of their party preference, their vote choice may be relatively impressionable to short-term factors such as political events or changeable evaluations of party leaders (Sevi, 2021). As Campbell and colleagues (1960: 497) put it: "For the young voter the Democratic Party is, for better or for worse, the party of Stevenson and Truman […] A political party signifies little more to them than its current leaders." Another short-term determinant of vote choice that is of particular interest to the current research question is the so-called 'bandwagon effect' (Robinson, 1937). Relatively uninfluenced by long-term determinants, young voters may be tempted to join the winning team by media coverage of which parties are gaining in the polls (Stolwijk and Schuck, 2019;Van der Meer et al., 2016).
The fourth reason why the youngest voters may be most volatile lies in generational differences in the determinants of vote choice. In most countries, voters who were born before the 1950s were socialized in an era of cleavage politics. Even many decades later, this generation may therefore still be most likely to base its vote choice on structural factors such as social class or religion (Franklin, 1992). Voters who grew up after the era of cleavage politics may contrarily be less constrained by long-term factors and hence more susceptible to short-term influences. Research on this hypothesis has however yielded mixed results. Some studies have indeed shown that long-term factors such as class and religion are more predictive of vote choice among older voters (Franklin, 1992;Maggini, 2016;Van der Brug, 2010;Walczak et al., 2012), but these results may alternatively be attributed to life-cycle differences. Age-period-cohort studies that estimated generational differences while controlling for life-cycle effects revealed that the vote choice of younger generations does not so much depend less on long-term factors (i.e., dealignment), but rather on different long-term factors (i.e., realignment) such as stable issue attitudes on new cultural issues (Rekker, 2016;Van der Brug and Rekker, 2021).
In sum, there are four reasons to expect that the youngest voters are most volatile and hence that winning parties gain disproportionately among this age group: (1) life-cycle differences in party loyalty, (2) generational differences in party loyalty, (3) life-cycle differences in the determinants of vote choice, and (4) generational differences in the determinants of vote choice. The purpose of the present study was not to empirically disentangle these mechanisms or to determine which offers the best explanation for its findings. Indeed, this is precisely where this investigation differs from previous studies. Whereas previous research has focused on unraveling the mechanisms that explain vote switching among individual citizens (i.e., individual-level volatility), the present study examined to what extent the unstable party preferences of young voters fuel electoral shifts between political parties (i.e., aggregate-level volatility). Drawing from the aforementioned theoretical accounts, the first hypothesis was postulated as follows: H1. Parties that win vote share gain disproportionately among young voters.

Young voters and new parties
Electoral volatility is not just driven by transfers of vote share between existing parties (i.e., alteration), but also by the entrance of new parties to the political arena (i.e., regeneration; Chiaramonte and Emanuele, 2017). There are compelling reasons why young voters may contribute not just to electoral shifts between existing parties, but also and perhaps even more so to the breakthrough of new parties. In their first election, new parties can by definition be seen as winning parties because their previous vote share was zero. All aforementioned reasons why winning parties may gain disproportionately among young voters therefore also apply to new parties. In addition, there are four reasons to expect that new parties may benefit even more from young voters than winning existing parties. The first reason can be found in life-cycle differences in dormant party loyalties. Although winning existing parties also attract voters who did not support them in the previous election, many of their 'new' supporters may have already cast their ballot for them at some point in their lives. Existing parties can therefore make electoral gains by winning back the support of citizens who already developed some loyalty to them in earlier elections and for whom they are still part of their 'consideration sets' (Oscarsson and Rosema, 2019). New parties however lack this dormant electoral potential and instead have to build their voter base from the ground up. Even more than winning existing parties, new parties therefore depend on citizens without party loyalties and such voters may be found mainly in the youngest age group.
A second reason why new parties may attract a young electorate lies in life-cycle differences in voters' preference for effective parties. A crucial challenge for every new party is to convince potential voters that it will be able to get elected, to express their views in debates, and ultimately to influence policies (Eatwell, 2003;Van der Brug et al., 2005). Research shows that voters often question the effectiveness of new parties and that they are therefore hesitant to give a new party their vote, even if they support its ideas (Bos and Van der Brug, 2010). Young people have however never observed the proven track record of established parties and may therefore be less likely to prefer existing parties over newcomers that have yet to prove their effectiveness. Goerres (2009: 71) for example reasoned that "some parties can leave a more prominent impression on electors through repeated government participation or their relatively constant electoral size" and that established parties therefore "seem to be more successful with older voters in electoral systems with proportional representation as they can repeatedly impress ageing voters." A third rationale to expect success of new parties among young voters involves generational differences in voters' issue priorities. New parties commonly attempt to break the dominance of established parties by mobilizing new issues such as immigration, the environment, or European unification (De Vries and Hobolt, 2020;Lucardie, 2000;Rochon, 1985). Such new issues are usually prioritized most by young citizens, whereas earlier generations continue to view politics in terms of older debates that were salient during their early life (e.g., Inglehart, 1977;Van der Brug and Rekker, 2021). Inglehart (1977) for example demonstrated that young baby boomers in the 1970s were more likely than earlier generations to prioritize new 'postmaterialist' issues such as environmental protection over 'materialist' issues such as economic growth. Likewise, Carmine and Stimson (1981: 107) argued that new issues "create the opportunity for political change" and that "new citizens are its most likely agents." Following this premise, they viewed new generations as a driving force behind the rise of racial issues that occurred in the United States since the 1960s. Research on electoral realignment in Western Europe has similarly revealed that citizens who came of age after about 1980 are most likely to base their political identity and vote choice on new cultural issues such as immigration or European unification (Rekker, 2016;Van der Brug and Rekker, 2021;Walczak et al., 2012).
The fourth reason why new parties may perform well among young voters lies in generational differences in media use. Most voters have no other way to learn about new parties than through media. Getting news media attention is therefore even more important for new parties than for existing parties, but it may also be more difficult for them (Art, 2006). News media will typically pay little attention to an unknown party that has yet to prove its relevance. Because new parties are more likely to pursue a radical agenda (e.g., the radical right), they also face a greater risk to be stigmatized as anti-democratic in traditional news media (Art, 2006;Van Spanje and Azrout, 2019b). Due to these obstacles, new parties may often fail to attract older voters who exclusively consume traditional offline news. Young generations are however much more likely to learn about politics through new online media (Newman et al., 2020). Social media may therefore offer new parties a unique opportunity to quickly become visible among young citizens without the interference of media practitioners.
To sum up, there are four reasons to expect that new parties benefit even more from young voters than winning existing parties: (1) lifecycle differences in dormant party loyalties, (2) life-cycle differences in voters' preference for effective parties, (3) generational differences in voters' issue priorities, and (4) generational differences in media use. Previous research has documented many instances of new parties that attracted a young electorate (e.g., Goerres, 2008;Maggini, 2016). However, the evidence is so far mostly anecdotal because research has never examined the success of new parties among young voters across a large number of countries and decades. Moreover, no study has yet compared the success of new parties among young citizens with the popularity of winning existing parties among this age group. It is therefore yet unclear if new parties do well among young voters as a result of their newness, or simply because they are by definition winning parties in their first election. The second hypothesis of this study was therefore formulated as follows: H2. New parties gain even more disproportionately among young voters than winning existing parties.

Young voters as trendsetters
By supporting new and winning parties, young people may be enthusiastic followers of electoral trends. But could they also be trendsetters who initiate electoral developments that eventually carry over to older voters? This question relates to a sociological debate about the role of youths in social change that extends far beyond the domain of political behavior (e.g., Meeus, 1992;Münchmeier, 1982;Prakke, 1959;Rogers, 1995). For example, Rogers (1995) argued in his influential work 'Diffusion of Innovations' that young people often function as 'innovators' or 'early adopters' in the diffusion of new ideas, behaviors, or products. Likewise, models of change in women's dress have commonly distinguished between 'trickle-down' processes in which youths follow the upper class and 'bottoms-up' processes in which young people initiate fashion trends that carry over to older citizens (Behling, 1985). A similar distinction can be made between three different models that describe the role of youths in political change. The first paradigm views young citizens as revolutionaries for change. This perspective describes a state of generational conflict in which young people strive for social change and clash with older citizens who want to keep society as it is. The most well-known example of this paradigm is probably Inglehart's (1977) seminal work 'The Silent Revolution,' in which he described the rise of a new generation that sought social change in the 1970s along the lines of new 'postmaterialist' values.
A second model views young citizens as acceptors of change. Like the previous perspective, this paradigm describes a conflictual relation between generations. Instead of viewing young people as revolutionaries, this paradigm however views older citizens as reactionaries who mobilize against a process of change that was already happening. As Delli Carpini (1989) put it: "It is the older generations that may be potentially the most revolutionary (or, more accurately, reactionary), since it is their norms that are most at odds with the realities of the society in which they live." Following this paradigm, Norris and Inglehart (2019) for example argued that the rise of authoritarian populism in the early 21st century was fueled by a 'cultural backlash' in which older generations mobilized to stop and reverse an ongoing process of progressive cultural change. In line with this perspective, research has shown that citizens who grew up with high levels of globalization, immigration, and European unification are considerably more supportive of these developments than earlier generations (Down and Wilson, 2013;Rekker, 2018). In another example, a study on generational differences in the United Kingdom revealed that citizens who came of age during or after Thatcher's administration are more accepting of her conservative reforms than older citizens (Grasso et al., 2019). The third paradigm views young people as a vanguard of change. Whereas the previous two models focused on generational conflict, this third perspective describes a more harmonious relation between young and old. In the vanguard model, young people are viewed as trendsetters who are the first to initiate, explore, or embrace new developments. Older citizens are however not viewed as rigid defenders of the status quo, but rather as curious observers and potential followers of the changes that young people initiate. In protests against authoritarian regimes, youths are for example often portrayed as a vanguard that leads the way for older protestors (Lührmann, 2015;Sayre and Yousef, 2016). Also in line with the vanguard perspective, age-period-cohort studies have documented instances in which generational differences became smaller over time because older citizens eventually started to resemble young generations (e.g., Rekker, 2016).
Drawing from this vanguard paradigm, the present study examined if young voters could function as trendsetters in elections, which would imply that electoral shifts among young voters predict similar changes among older citizens in the next election. There are three reason to expect that this could be the case. The first reason can be found in the idea of trickle-up political socialization. After young voters have embraced a new or winning party, they may to some extent be able to influence older voters through direct personal interactions. Although such trickleup political socialization may occur wherever young and old meet, the most likely place is within the family. Many young people still live with their parents when they are first allowed to vote and they commonly discuss their party preference at home (Hooghe and Stiers, 2020). Although the literature has long viewed political socialization as a unidirectional process of parental transmission to children (e.g., Jennings and Niemi, 1968), more recent studies have revealed that young voters can also influence their parents' attitudes and behavior (Dahlgaard, 2018;McDevitt and Chaffee, 2002;Wong and Tseng, 2008).
A second way in which youths may influence older voters is by creating reputational boosts for parties. Achieving electoral success among young voters may make a party look energetic, promising, and futureoriented (e.g., Maclean's, 2015). The British Labour leader Jeremy Corbyn for example received quite some favorable media coverage for his success among young voters in 2017 (e.g., The Guardian, 2017). Indeed, parties often seem to emphasize the youthfulness of their electorate by for example surrounding their leaders with young supporters during public appearances (New York Magazine, 2012). Such reputational boosts from young voters may be especially important for new parties that still need to prove their viability. As argued above, older voters may be more hesitant to vote for a new party that has yet to prove its effectiveness (Goerres, 2009). Young voters may therefore be vital for a new party's initial breakthrough in its first election. After a new party has proven its viability by getting elected, older voters may follow in its second election. In such cases, young voters may set in motion a process of party system change that eventually forces older voters to adapt to a new political reality (Franklin and Van Spanje, 2012).
The third reason why a party's popularity among young people in one election may predict success among older voters in the next lies in the process of generational replacement. During their first elections, young voters develop voting habits and party attachments that can remain rather stable once established (e.g., Dinas, 2014). Many young voters who voted for a new or winning party may therefore remain loyal to that party in the next election and possibly beyond. In other words, a party's success among first-time voters in one election may predict success among second-time voters in the next election simply because this is the same cohort. In line with this idea, research on generational differences in party preference has revealed that many older voters still support the party that was most popular when they were first allowed to vote (Tilley and Evans, 2014).
In sum, there are three reasons to expect that electoral shifts among young voters are predictive of similar changes among older citizens: (1) trickle-up political socialization, (2) reputational boosts for parties, and (3) generational replacement. Nonetheless, no previous study has examined this possibility. The third hypothesis of this study was therefore formulated as follows: H3. Electoral shifts among young voters predict similar changes among older citizens in the subsequent election.

Data
This study used a new dataset that was created by compiling vote share estimates from existing national election surveys (see Table 1 for an overview). The rows in this dataset represent 1,232 party-election combinations, while the columns contain the percentage vote shares that a party obtained among respectively (1) all respondents of the election survey (N = 557,132), (2) respondents under age 25 (N = 47,993), and (3) respondents over age 24 (N = 509,139). The cut-off between young and older voters was age 25 because research on attitude formation indicates that this is roughly the age at which political learning slows down and voters' political orientations approach adult levels of crystallization and stability (e.g., Rekker et al., 2019). The youngest respondents had just reached the legal voting age, which was 18 in the vast majority of analyzed elections with some exceptions at 16 and 21. This study's dataset had to be compiled with unweighted vote share estimates because weights were not available for every survey.
The analyzed election surveys were conducted between 1948 and 2019 in 21 countries: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, the Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, the United Kingdom, and the United States. The respondents of these surveys were interviewed after all first-order elections: the presidential elections in (semi-)presidential systems (i.e., France and the United States) and the parliamentary elections in parliamentary systems (i.e., all other examined countries). All survey years that were available as of June 2020 were included in the dataset. In the case of missing survey years, only the elections from after the interruption were incorporated to provide an uninterrupted time series. This study used all available election surveys from Western established democracies, but not from new (i.e., post-1989) or non-Western democracies. This demarcation was made for a theoretical and a methodological reason. Theoretically, the rationale for this study's hypotheses draws largely from accounts of electoral change in Western established democracies. Although some of these accounts may generalize more broadly, others may not. Methodologically, expanding this study to new or non-Western democracies would be problematic due to a lack of available election surveys from such countries that go back far enough in time to allow for reliable estimates.
This study examined the vote shares of 165 political parties. Without any inclusion criteria, a disproportionate share of the data would have consisted of rather insignificant parties. To avoid this convolution, a party was only included if two conditions were satisfied: (1) that the party had participated in at least two elections and (2) that it had obtained a vote share of at least five percent in at least one election. For every party-election combination, a distinction was made between three types of parties. The first category consisted of recurring parties. A party was classified in this category if the same organizational entity had already participated in the previous election, either under the same name or under a different name (i.e., successor parties). For this type of parties, electoral gains could be calculated by comparing a party's vote share with its result in the previous election. A second category consisted of new parties. A party was classified in this category if it had not participated in the previous national election either under the same name, under a different name, or as a combination of several parties. This implies that fission parties, but not fusion and successor parties, were viewed as new parties in this study (as proposed by Hug, 2001). National parties that were formed as a merger of local parties (e.g., Lega Nord in Italy) were however also classified as new parties because their predecessors did not participate in national elections. Based on this definition, this study could treat the entire vote share of a new party as an electoral gain because its vote share in the previous election was zero. Fusion parties and electoral alliances finally made up a third category. The vote share of a merger or alliance in its first election could not be used to calculate electoral gains because this vote share could not be compared to a result from the previous election (like for recurring parties), while it could also not be defined as zero (like for new parties).

Variables
The primary outcome variable in this study was the hypothesized youth bonus, which indicates to what extent a winning party gained disproportionately among young voters. Based on a party's percentage vote share among young (age <25) and older voters (age >24), this variable could be calculated as follows for a given party (i) in a given election (t): A party's score on this variable therefore indicates the percentage point difference between its electoral gains among young and older voters. Because the hypotheses focused on winning parties, the youth bonus was only calculated for instances in which parties' overall vote share increased: Subset: The youth bonus variable was used for testing the hypothesis that winning parties gain disproportionately among young voters (H1) and the hypothesis that this effect is stronger for new parties than for existing parties (H2). For a further examination of the hypothesis on new parties, a second outcome variable was the age gap in a party's vote share, which indicates the percentage point difference between its vote share among young and older voters in the same election: The third outcome variable indicates the percentage point change from the previous election to the current election in a party's vote share among older voters: Similarly, this study included two independent variables that indicate the percentage point change from the election before the previous election to the previous election in a party's vote share among respectively young voters and voters of all age groups: The variables in equation (3) through (5) were used to test the third hypothesis that electoral shifts among young voters predict similar changes among older citizens in the subsequent election (H3). As a control variable, this study finally included a dichotomous ideological classification of political parties as either 'left-wing' or 'right-wing.' A party's score on this variable was based on where its voters placed themselves on a left-right scale. A party was classified as left-wing if survey respondents who had voted for this party on average placed themselves below the center of a left-right scale, whereas it was classified as right-wing if its voters placed themselves above the center of the scale. This study used voters' ideological self-placement, rather than their placement of parties, due to better data availability. Using respondents' ideological placement of the parties would however have resulted in nearly identical classifications and, for all intents and purposes, identical results.

Estimation
The analyses were conducted using ordinary least squares (OLS) estimation with standard errors that were robust to heteroscedasticity and clustering within elections. By estimating cluster-robust standard errors, this study could account for the fact that (changes in) the vote shares of different parties within the same election are not independent observations. The fact that the electoral gains of winning parties equal the losses of losing parties was furthermore accounted for by calculating the youth bonus variable only for winning parties. If winning parties gain disproportionately among young voters, this logically implies that losing parties also lose disproportionately among this age group. By only including the vote share of winning parties in these analyses, this study could avoid analyzing every electoral shift twice (i.e., as a gain and as a loss).
Another methodological issue that has to be considered is the role of measurement error in the estimation. This study analyzed estimates of parties' vote share in different age groups that were obtained from election surveys. The sampling error of these surveys therefore manifests itself as measurement error in the vote share estimates. On average, every election survey interviewed 219 young (age <25) and 2,325 older respondents (age >24). Although these average numbers suffice for a satisfactory reliability of vote share estimates, there were also instances in which reliability suffered because the vote share of a small party had to be estimated with a relatively small number of young respondents (e. g., N = 50). Because vote share estimates featured as the dependent variable for the first and the second hypothesis, this measurement error can only have produced error, but not bias, in these instances (King et al., 1994). In other words, the errors cancelled each other out across all 1,232 party-election combinations to produce an unbiased estimate of the mean youth bonus. The issue of measurement error did however make it difficult to estimate the mean youth bonus separately for each individual country, especially for countries with a limited number of election surveys (e.g., Austria or Greece). Although this study included an exploratory analysis on cross-country differences, all hypotheses were therefore tested across countries. For the third hypothesis, measurement error may additionally have produced a downward (but not upward) bias in the estimates because vote shares among young voters were used as the independent variable in this analysis (King et al., 1994). The estimated effect size for the hypothesis on trendsetting should therefore be seen as a conservative estimate of the true value of this effect.

Young voters and winning parties
The first hypothesis was tested by analyzing all instances of electoral gains with an intercept-only model (model 1) that featured the youth bonus (equation (1)) as dependent variable. As displayed in Table 2, this analysis revealed a significant mean youth bonus of 0.60. Confirming the hypothesis, this implies that winning parties on average gained 0.60 percentage points more among young voters (5.08%) than among older citizens (4.48%). This means that a winning party typically gained 1.13 percent (5.08/4.48) among young voters for every percent that it gained among older voters. This pattern is depicted in Fig. 1 for fifteen parties that revealed a particularly pronounced youth bonus effect. The Icelandic pirate party (Píratar) for example obtained more votes among young people when it was first elected to parliament in 2013, it then gained disproportionately among young voters when it won vote share in 2016, but it also lost more among youths when its vote share decreased in 2017. Likewise, the Dutch social liberals (Democraten 66) Table 2 Regression models for the main analyses. Shift in previous election among young voters (Equation (4)) 0.13 (0.06)* Shift in previous election among all voters (Equation (5)) − 0.36 (0.08)*** Model Observations 952 R 2 6.4% Note. Estimates with standard errors in parentheses.+p < .10; *p < .05; **p < .01; ***p < .001. Table 2 and Fig. 2 for the average effect size).

R. Rekker
Electoral Studies 75 (2022) 102425 8 gained disproportionately among young voters during its three electoral peaks in 1981, 1994, and 2017. Nonetheless, the overall youth bonus effect size of 0.60 percentage points can be considered surprisingly modest.
As an exploratory analysis, this study furthermore compared the mean youth bonus across time and countries. As displayed in Appendix 1, this analysis revealed a clear absence of systematic over-time variation in the youth bonus effect. The pattern across countries is however less clear. Although the estimates clearly differ between countries, these differences are only slightly larger than what could be expected based on random variation. Although the F-test on country effects revealed a significant p-value, only the effect for New Zealand differs significantly from that of Sweden, which was chosen as the reference category due to its large sample size and because the estimate for this country was closest to the cross-country average. This exploratory analysis therefore failed to provide clear evidence for either the existence or the absence of cross-country differences in the mean youth bonus. As a robustness check, Appendix 2 shows that all results of this study are identical when cross-country variation is controlled for by adding country dummies to the regression models.

Young voters and new parties
To test the second hypothesis, another regression model (model 2) added a dummy variable that distinguished new parties (i.e., in their first election; N = 48) from existing parties (i.e., all other party-election combinations; N = 1,184). As hypothesized, the results (model 2) revealed that the mean youth bonus was larger for new parties than for existing parties. As depicted in the top-left corner of Fig. 2, the effect size was very strong for this second hypothesis. New parties on average gained 2.24 percentage points more among youths (9.33%) than among older voters (7.09%), which was a five times larger effect than the modest (but still significant) mean youth bonus of 0.44 percentage points for existing parties. This means that a new party on average gained 1.32 percent (9.33/7.09) among young voters for every percent that it gained among older voters.
To provide a more thorough examination of the second hypothesis, this study furthermore estimated three regression models with the age Fig. 2. The mean youth bonus (top-left corner) and electoral age gap (top-right corner) for new and existing parties with a 95% confidence interval and the vote shares of the ten new parties that won most disproportionately among young voters (bottom half). gap in parties' vote share (equation (2)) as dependent variable. The first of these models (model 3) revealed that the average vote share of new parties was 2.15 percentage points higher among young voters than among older citizens, whereas existing parties reversely obtained 0.44 percentage points less among this age group. A second model (model 4) added a dummy variable that indicated a party's ideological position. The difference between new and existing parties was still significant after controlling for ideology and even increased from 2.58 to 3.05 percentage points. This means that the success of new parties among young voters cannot be attributed to any particular ideology. Interestingly, this model also revealed that left-wing parties obtained a higher average vote share among youths than among older voters (+1.41 pp), whereas the opposite was true for right-wing parties (− 2.29 pp). Although outside the scope of this study, this finding is consistent with previous research on life-cycle and generational differences in vote choice (e.g., Tilley and Evans, 2014). A third model (model 5) finally added an interaction term between newness and ideology. This interaction effect was small and non-significant. As depicted in the top-right corner of Fig. 2, this means that new parties on both sides of the political spectrum enjoy an equally large electoral advantage among young voters. The bottom-half of Fig. 2 depicts the ten new parties that revealed the largest age gap in their vote share. The most striking example is the Italian Movimento 5 Stelle. During its first election in 2013, Beppe Grillo's party obtained almost double the vote share among young citizens (45%) compared to its result among older voters (23%).
The other examples are an ideologically diverse group of new parties ranging from the Bloc Québécois in Canada to Bündnis 90/Die Grünen in Germany and Lega Nord in Italy.
As an exploratory analysis, this study also examined to what extent new parties can maintain their advantage among young voters after their first election. This analysis added two dummy variables to model 3 that indicate new parties in their second and third election. The results revealed that the age gap in vote share differs most from older parties for new parties in their first election (b = 2.70; SE = 0.77; p < .001) and that this effect declines in their second (b = 1.75; SE = 0.77; p = .023) and third election (b = 1.45; SE = 0.74; p = .052). In line with the theoretical Fig. 3. Top half: Predicted electoral gains among older voters as a function of electoral gains among young voters in the previous election with a 95% confidence interval (i.e., estimates from Model 6). Bottom half: Six of the most pronounced instances in which electoral gains among young voters were predictive of similar shift among older voters.
reasoning of this study, this pattern indicates that new parties enjoy the largest relative popularity among young voters in their first election and that they may gradually lose this advantage in subsequent elections.

Young voters as trendsetters
The third hypothesis was finally tested by estimating a regression model (model 6) with as dependent variable parties' shifts in vote share among older voters in the current election (equation (3)). The independent variables were shifts in the previous election among young voters (equation (4)) and among the entire electorate (equation (5)). The latter variable was included to control for the fact that parties that win an unusually large vote share (e.g., due to strategic voting) often see a downward correction in the next election. Confirming the hypothesis, the results revealed that a percentage point gain or loss among young voters predicts a shift of 0.13 percentage points in the same direction among older voters in the next election. This effect is depicted in Fig. 3 for both the average pattern (top half) and some of the most pronounced instances (bottom half). The Movimento 5 Stelle is again an interesting example. While the initial electoral breakthrough of this party in 2013 was fueled largely by young voters, older voters followed in the same numbers during its second election in 2018.

Discussion
In the footsteps of Karl Mannheim (1928), many scholars have argued that social change occurs first and foremost among young citizens (e.g., Prakke, 1959). This study examined to what extent this proposition applies to electoral shifts by comparing parties' electoral gains among young and older voters in 21 countries between 1948 and 2019. The results first of all showed that young voters are indeed enthusiastic followers of electoral trends and that winning parties therefore gain disproportionately among this age group. This article introduced the term youth bonus to refer to this phenomenon. Of course, the additional gains of winning parties logically imply that losing parties also lose disproportionately among youths. Previous research has already demonstrated that individual young citizens are more likely to switch their vote than older citizens (e.g., Jennings and Markus, 1984), but the present study was the first to demonstrate that this individual-level volatility indeed manifests itself in actual transfers of vote share between political parties (i.e., aggregate-level volatility). Whenever a winning party seems particularly popular among young voters, the youth bonus effect should therefore be taken into account as a potential explanation for this popularity. This study however also revealed that the effect size of the youth bonus is surprisingly modest in the light of the extensive literature on the instability of young voters' political preferences. On average, winning parties gained 1.13 percent among young voters for every percent that they gained among older voters. Although meaningful, this youth bonus effect only provides a small piece of the puzzle when explaining why some parties attract a much younger electorate than others. Indeed, accounts that use generational differences in core political attitudes to explain the electoral behavior of young citizens (e.g., Norris and Inglehart, 2019) explain a much larger share of the relevant variation.
This article outlined four theoretical mechanisms that could drive the youth bonus effect: (1) life-cycle differences in party loyalty, (2) generational differences in party loyalty, (3) life-cycle differences in the determinants of vote choice, and (4) generational differences in the determinants of vote choice. Although it was not an aim of this study to empirically distinguish between these four processes, the clear lack of over-time differences in the youth bonus effect may nonetheless shed some light on this issue. If vote switching would have been a characteristic of any particular generation, this study would probably have observed more over-time variation between the 1950s and the 2010s. The constancy of the youth bonus effect over time is therefore consistent with previous research that attributed the instability of young people's electoral behavior to life-cycle effects, rather than generational differences (e.g., Dassonneville, 2013).
Although the magnitude of the youth bonus seems limited for winning existing parties, this study also revealed that this effect is no less than five times larger for new parties. In their first election, new parties on average obtained a 1.32 times larger vote share among youths than among older citizens. This relative advantage among young voters moreover applied equally to new parties on both sides of the political spectrum. This article described four mechanism that could explain this phenomenon: (1) life-cycle differences in dormant party loyalties, (2) life-cycle differences in voters' preference for effective parties, (3) generational differences in voters' issue priorities, and (4) generational differences in media use. Taken together, these four mechanisms describe how the rise of new parties could be fueled by an interplay of new issues, new media, and new voters. Although many previous studies have provided anecdotal evidence for the success of new parties among young voters (e.g., Maggini, 2016), the present study was the first to systematically compare and quantify this effect across a large number of countries and decades.
By supporting new and winning parties, young voters follow and strengthen electoral trends. Nonetheless, youths can only be seen as 'trendsetters' if their electoral behavior eventually carries over to older voters. This study provided evidence for this idea by demonstrating that electoral shifts among young voters indeed predict similar changes among older citizens in de subsequent election. This article proposed three mechanisms that could explain this pattern: (1) trickle-up political socialization, (2) reputational boosts for parties, and (3) generational replacement. The finding that electoral shifts among young voters can carry over to older citizens indicates that the impact of young voters on electoral volatility may be larger than their relatively small numbers (e. g., about nine percent of respondents in the analyzed election surveys) suggest.
A limitation of this study is that it failed to provide evidence for either the existence or the absence of cross-country differences. Although the effect size of the youth bonus was modest when averaged across countries, this study's exploratory analysis hinted that it could be substantially larger in some countries (e.g., the Netherlands and the United Kingdom) than in others. Examining cross-country variation in the youth bonus effect therefore remains a challenge for future research. Because the present study prioritized parsimony by only distinguishing between voters under and over the age of 25, future research could furthermore expand upon its analyses by comparing a larger variety of age groups. More research is also needed on this study's findings on trendsetting. The present study was the first to demonstrate that the electoral behavior of youths may carry over to older voters and this effect was only significant at the 0.05 level. Some caution is therefore warranted until future studies establish the robustness and explanatory mechanisms of this finding. A final note of caution is that this study analyzed vote share estimates that were obtained from unweighted survey data because weights were not available for all election surveys. The results of this study could be influenced if the assumption is violated that any sampling or response survey bias in volatility estimates (such as a tendency to underreport support for losing parties) affect young and older voters to roughly the same extent.
Taken together, this study's findings point out that young voters are important drivers of electoral volatility. Youths strengthen electoral shifts by supporting winning parties and they importantly facilitate the breakthrough of new parties. Moreover, there are instances in which young people initiate electoral trends that eventually carry over to older voters, either because youths replace older voters or because they influence them. Young voters may therefore be seen as a vanguard of electoral volatility that can lead the way for older citizens by being the first to explore and embrace new electoral developments.

Funding details
This research was supported by the Dutch Research Council (NWO) with a VENI grant awarded to Dr. Roderik Rekker. Grant Number: VI. Veni.191R.039 'Are millennials transforming politics? A study on generational differences in voting.'

Data availability
Data will be made available on request.

Declaration of competing interest
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.