Age gaps in political representation: Comparing local and national elections

.


Introduction
The study of young adults' political representation is a burgeoning field.Scholars have examined the descriptive representation of young people around the world, focusing on assessing the extent to which young people are underrepresented (Belschner and Garcia de Paredes, 2021;Joshi, 2013;Stockemer andSundström 2019, 2022b), as well as the macro-level mechanisms that contribute to disparities in age representation (Krook and Nugent, 2018;Stockemer and Sundström, 2018).While there have been significant strides in growing this area of scholarship within representation studies, there is little comparative research about young candidates' fate in processes of political selection and election.In this article, we compare the role of candidate age in local and national elections.We address a gap in the existing youth representation literature that, so far, has done little in the way of investigating questions concerning where, or at which level, young people tend to seek office, and at which level these young people are nominated as candidates and might be successful.Moreover, when and where research has investigated the selection of young candidates, it has tended to do so for a single level of government and mainly in the context of the United States (Lawless and Fox, 2015;Shames, 2017;Stockemer et al., 2023).
Our article also contributes evidence about the role of candidate age in sub-national politics.There are at least two reasons why this area of research has remained relatively minor within the literature on youth representation: data availability and predeterminate assumptions.First, data for youth representation at the sub-national level is scarce.While there are numerous efforts to gauge young people's descriptive representation at the national level by the Inter-Parliamentary Union, and the Worldwide Age Representation in Parliaments (WARP) datasets (Stockemer and Sundström, 2022a), age-data, specifically on the candidate level, remains rather difficult to access.Second, like in the early literature on women's representation, from which most of the youth literature has followed theoretically, there has been an unspoken assumption that political careers for young people are most likely to begin in local politics.
In this article, we analyze the role of age in candidate selection and election, the former pertaining to the presence of young candidates on party lists and latter to their success in becoming elected representatives.We compare these dynamics between local and national elections.Our empirical analysis focuses on the case of Finland, an ideal context for comparing the role of candidates' age in local and national elections due to its open-list proportional representation system that is similarly applied in both elections.In this system, parties select candidates to run on their electoral lists.It is mandatory for voters to cast a preference vote for a single candidate in their electoral district.The proportional share of seats a party list has secured are then distributed to the candidates who won most votesthey get elected.Since the lists are open and candidates appear in alphabetical order, parties cannot predetermine which candidates are elected by placing them in high or low positions on the list.Hence, the Finnish case furthermore allows for direct observation of how appealing voters find young candidates and provides insights into the political careers of young aspirants in an open-list proportional representation system.
In our analysis, we compare the effect of running in local versus national elections on two main outcomes: first, in the age of candidates who stand for candidacy, and second, in their electoral success.Addressing dynamics of candidate supply and demand, we test two competing hypotheses about how age affects selection and election differently in local vs.national elections.The "stepping stone" hypothesis assumes that it is easier for young candidates to be selected and elected in local elections mostly due to higher party demand.In contrast, the "parachute" hypothesis suggests that the supply of young candidates is higher for national than for local elections, and that the most promising young aspirants will find it easier to be selected and elected at the national level.While we do not attempt to make a causal argument about differences between the levels, we suggest that comparing them sheds light onto how level-specific voter and party behavior shape the start of political careers.
We conduct two main empirical analyses.First, we draw on detailed candidate register data including all candidates that have stood in six municipal and national elections between 2011 and 2021 (over 100,000 candidates in total).We employ linear probability models to estimate how the level of an election correlates with the age of selected candidates and the relative impact of age on electoral performance.Our findings provide slightly more support for the parachuting hypothesis, specifically at the selection stage.The analyses reveal that while national elections are more competitive, comparatively more young candidates are running in these contests than at the municipal level.We also find that the electoral disadvantage for young candidates is slightly larger in municipal than in national elections, and show that the main source of age-related disadvantages is a lack of incumbency.In the second analysis, we further explore potential explanations for these findings.Using election survey data, we show that older voters place more emphasis on candidate experience and were more likely to vote for a same-age candidate in municipal compared to national elections.As senior voters represent a large and growing share of electorates, their preferences weigh more heavily than those of young voters.This, we argue, can help explain the differences between the levels discovered in our first analysis.
The article proceeds as follows.We begin by outlining previous scholarship and build our theory about how the level of government affects youth candidacies and representation.We draw on the research on women's political representation, although we recognize that women's experiences with political marginalization are different from that of aged-based marginalization.However, much of the scholarship on young adult political underrepresentation started by exploring factors shown to impact women's underrepresentation and has demonstrated continuities and differences between the mechanisms that impact these groups' underrepresentation.After discussing theory, we put forward two hypotheses before discussing the Finnish context.We then present our data and methods, display the results, and discuss our findings.

Literature review
To understand the composition of our representative bodies, we must look closely at the process through which individuals become recruited into politics.Scholars have often adopted a four-stage model of political recruitment (Norris and Lovenduski, 1995).The model starts with a large number of citizens who are eligible to run for political office and moves to those who not only are eligible but also aspire to seek office.This aspirant pool is whittled down to a handful of individuals who actually secure a nomination.The final stage produces the smallest group: those who are ultimately elected to office.Within this framework, scholars have often examined "supply-side explanations" which address why some individuals are willing to run for office, and "demand-side explanations" which focus on why certain people are selected as candidates and eventually elected as representatives.
Although most European electoral systems now permit voters to show preference for individual candidates rather than just political parties, the role of parties in recruiting and selecting candidates remains central (Renwick and Pilet, 2016;Vandeleene, 2023).However, the lines between recruitment and selection and between candidate supply and party demand can sometimes blur.As Siavelis and Morgenstein argue, "it is often the same elites that cultivate, identify, and name candidates, initiating a process that culminates in candidate selection" (Siavelis and Morgenstern, 2008: 31).For instance, when political parties find themselves with a dearth of potential candidates for upcoming electionsas they oftentimes do at the local elections (Kjaer, 2007;Sipinen, 2021) they may have to proactively sift through their networks to motivate qualified individuals to engage in the electoral process.
Our analysis focuses on identifying where, or at which level, young candidates are most likely to secure a position on the party list and achieve electoral success.We employ the terms "candidate selection" and "election".The former refers to a combination of candidate supply and party demand, while the latter represents a combination of party supply and voter demand.We do not disentangle the causes for the selection of candidates to the party lists, nor do we test what leads voters to (not) select young candidates.Rather, we focus on demonstrating where young people are most likely to appear as candidates and where they are successful.

The selection of young candidates
From the supply-side perspective, a first key question is on which level young candidates are more likely to come forward as candidates.Contrary to scholarship which has argued that young people are disinterested in electoral politics (Farthing, 2010;Henn and Foard, 2012;Kitanova, 2020), more recent scholarship has demonstrated that young adults' participation in party politics, particularly through political party youth wings, is on the rise, bucking trends of youth political party participation decline (Bolin et al., 2023;de Roon, 2020).These are most-likely politicians to seek political office (Stockemer and Sundström, 2023), but two major life-cycle factors impact their decisions to step forward as candidates (Fjellman and Rosén Sundström, 2021;Shames, 2017).First, young adults may lack the time and financial resources needed to seek political office (Belschner, 2023;Shames, 2017).Second, young adults are often in a period of transition which can force them to migrate away from their home to new cities/towns where they may have less social and political capital needed to foment candidacy (Flanagan and Levine, 2010;Krishna, 2002;Perkins and Long, 2002).
There are at least three reasons that young people might be more likely to step forward as candidates for the municipal level than the national level.First, local electoral contests are generally less resourceintensive than national elections; both with regards to political capital and monetary capital (Gavan-Koop and Smith, 2008;Gidengil and Vengroff, 1997;Kushner et al., 1997).Moreover, as young adults are most likely to be in education, or just starting a family, local politics can provide a means for young aspirants to maintain their educational and familial responsibilities while holding political office (Gidengil and Vengroff, 1997).Finally, young adults may view local election candidacies as a political stepping stone, offering them a 'political education' and means to accrue political capital and experience (Sharpe, 1970).
Yet, young people may also be more inclined to seek higher level office.First, there is a greater prestige associated with higher level positions that can provide enhanced career opportunities which may be particularly appealing to young adults who have longer time horizons and are at the start of their career (both within and external to politics) (Hooghe et al., 2004).Second, the digital age has lowered the financial obstacle associated with higher level office campaigns (Anlar, 2020;Earl and Kimport, 2011).Additionally, research suggests that young people can obtain greater political capital through their placement on party lists.Participating in elections enables them to build a loyal following in their districts over time, reaping the benefits of repeated-even if unsuccessful-candidacies (Söderlund and von Schoultz, 2023).
In addition to the decisions to step forward as candidates-supplyside arguments, political parties must also have a demand for your candidates, seeking their nominations and placing them on the ballot.As such, the second key question is whether political parties are more inclined to nominate young candidates for local offices, which are often considered less prestigious, or if they consider them equally viable for national seats.
Several theories such as the 'law of increasing disproportion' (Putnam, 1976), 'the higher the fewer' (Bashevkin, 1993), and the 'law of minority attrition' (Taagepera, 1994) posit that minority representation tends to wane at higher levels of the political hierarchy as parties and voters might prioritize candidates with more experience for these more prestigious and professionalized positions (Best and Cotta, 2000;Cotta and Heinrich, 2007).However, political parties, acting as rational prospectors, conduct comprehensive cost-benefit analyses to optimize their chances of securing the most seats in legislative bodies (Gallagher and Marsh, 1988;Vandeleene, 2023), which often relies on approximating how voters will perceive the list, incentivizing parties to find a balance between experience and desired candidate characteristics.
Scholars have demonstrated that proportional representation systems tend to elect more minority candidates as parties are incentivized to balance their tickets (Matland and Studlar, 1996).This is particularly true in open-list proportional representation systems (OLPR) where voters have the option to cast preference votes for individual candidates (von Schoultz, 2018;Shugart et al., 2005).With the decline of class-based voting and the erosion of strong party-voter ties, parties are compelled to offer diverse candidate lists that can appeal to a broad electorate.Hence, scholars contend that PR systems, in general, but particularly as OLPR, encourage the creation of 'balanced lists,' especially in terms of candidates' socio-demographic characteristics like age, gender, occupation, and residence (Arter, 2013;Gallagher and Marsh, 1988;Joshi, 2013;Krook and Nugent, 2018;Sundström and Stockemer, 2021).Given this, parties have incentives to select candidates at both the national and the local level.However, the supply of young candidates may challenge parties' abilities to create balanced lists at both levels.
At the local level, parties are likely to encourage the candidacy of young people to attract young voters, although they may have difficulty finding young candidates willing to come forward.Yet, when young candidates are found, they are often automatically added to the party list to fulfill the parties' aim of offering a balanced ticket.At the national level, the fewer available ballot slots increase competition.Despite this, the need for a 'balanced list' often results in some seats going to the most promising young candidates, potentially enhancing their representation.
The presence of young people on candidate lists is influenced by both supply and demand.While this paper does not delve into the mechanisms that contribute to the supply of young candidates, or their selection to party lists, we recognize that these processes are interconnected.Supply and demand factors thus inform our hypotheses about the level at which to expect more young candidates.

The election of young candidates
At which level are young candidates most likely to actually win elections?Are voters more likely to vote for young people at the local level where the stakes are perceived as lower, or are they willing to entrust them with national responsibilities?Unlike gender, age is also a proxy for life experiences, meaning that voters of all ages may view a younger age as indicative of a lack of competence for higher office, although this perception may be less pronounced at the local level, or, as Roberts and Wolak (2022) find in the United States, it may not significantly impact voters' preferences for these candidates.
While conventional wisdom about women suggests that marginalized groups may find greater electoral success at local level (Darcy et al., 1994;Gidengil and Vengroff, 1997), there is also a growing body of evidence questioning such linear paths in political careers (Fortin-Rittberger et al., 2019;Kjaer, 2010;Tolley, 2011).For example, Kjaer (2010) found that several countries demonstrated either equal representation between local and national levels (Italy, Portugal, and Romania), or an overrepresentation of women at the national level in comparison to the subnational level (Finland, Germany, Sweden and six others).The latter has been found specifically in cases with federalist and open-list electoral systems.For example, Fortin-Rittberger et al. ( 2019) demonstrate that in Germany women are more often represented at the national level than the local level.This is attributed to the fact that left-wing parties, which are more likely to nominate women, perform better nationally than they do locally.If young candidates are not evenly distributed across different political parties, a similar pattern could also account for the level of youth representation.
Age is also a proxy for generational attitudes, meaning that voters may intuitively feel that a candidate of a similar age would share their worldviews.The so-called 'affinity voting' thesis postulates that voters invoke baseline preferences for candidates based on shared sociodemographic characteristics (Dolan, 2018;McGregor et al., 2017).Sevi (2021), for instance, has demonstrated that on a global scale, age groups tend to vote for candidates within the age cohorts they belong (Sevi, 2021;Webster and Pierce, 2019).This is similar to scholarship on candidate sex which finds that there is in-group affinity for women candidates by women voters (Dolan, 2005;Erzeel and Caluwaerts, 2015;Helimäki et al., 2023).This suggests that voters see age as an important factor, which may have implications for their preferences for younger or older candidates.
Previous research indicates that the use of heuristics and, thereby, affinity voting is more common in elections where voters have limited information and face higher informational costs (Crowder-Meyer et al., 2020).This suggests that affinity voting is likely to be more prevalent at the local level, where a larger pool of candidates usually exists.If voters across all age groups are inclined to vote for candidates who are close in age to themselves, the age composition of the electorate is likely to be more directly mirrored in local elected officials compared to national ones.However, if younger people are less likely to vote in local, second-order elections, compared to national elections, then the higher average age of the electorate in local elections could actually work against young candidates.

Hypotheses
Based on the above literature, we formulate two types of hypotheses that relate to young people's selection as candidates and their election as representatives at the local vs.national level (Table 1).The hypotheses that anticipate higher youth candidacy and success at the local level are labelled "stepping stone" hypotheses, suggesting that local level politics is the "natural" starting point for political careers.The competing hypotheses, suggesting higher probability of selection and election at the national level, are labelled as "parachute", meaning that contrary to the stepping stone pathway, young representatives "fly over" the traditional pathways to higher office.The aim of these hypotheses are to frame this exploratory analysis.Rather than to tell a causal story concerning the causes of candidate emergence, we seek to demonstrate at which level young people are present on candidate lists, and where they are ultimately successful.

The Finnish context
The electoral system used in Finlandboth at parliamentary and municipal electionsis an open-list proportional representation (OLPR) system (for an overview, see von Schoultz, 2018).It is mandatory for voters to cast a preference vote for a single candidate in their electoral district, and voters can choose between numerous candidates of different age categories from multiple party lists.The number of votes each party gains is dependent on their candidates' personal votes, which at the end of the election are pooled together.This determines how many seats in the parliament or in municipal council each party gets.Within parties, the rank order by which candidates are elected is determined by their personal votes; those who have received most votes are elected.This means that parties cannot predetermine the rank order, as the voters decide instead.In addition, candidates are usually represented in alphabetical order on the ballot (Arter, 2021), meaning that voters are left without any cues of parties' order of preference.In Finland, therefore, parties cannot place candidates from disadvantaged social groups (such as the young) at the bottom of the list just to give an impression of diversity without real chances of being elected.
In Finland, parties nominate nearly all candidates in the municipal and parliamentary elections, leaving only few independent candidates.During campaigns, parties typically display pictorial advertisements featuring all their candidates.These "candidate galleries" that are generally placed in city or town centers enable voters to gauge candidates' ages from photos.This means that candidate age is easily mobilized as a heuristic.
Municipal elections are run separately from parliamentary elections in all municipalities in Mainland Finland (in 2023 the number of municipalities was 293), and parties can nominate 1.5 times as many candidates as there are seats in the municipal council.Council size varies between 13 and 85 depending on the population of the municipality, which means that the theoretical maximum of how many candidates each party can nominate varies between 19 in the smallest municipalities and 127 in the capital city of Helsinki.
Parties are motivated to nominate as many candidates as the system allows or they can find, because each candidate in an otherwise empty seat contributes additional votes.In municipal elections, parties aim to achieve full lists but often fail in many municipalities.To address this, parties actively persuade residents to run (Sipinen, 2021, 209-213).At the national level, there is still intense competition for nominations, although less so than during its peak in the 1950s-1970s (Arter, 2021).Unlike in parliamentary elections, where nomination requires, for example, name recognition, an established status within their own party, and substantial campaign funds (Arter, 2021), municipal-level candidacy has much lower barriers, at least outside the largest cities.In fact, many local candidates are "top-up candidates" who aim merely to attract votes for their party (Sipinen, 2021, 210).In terms of selection, research shows that in municipal elections, over half of residents tend to vote for candidates they know personally or through their social circle (Borg, 2017: 130-131).In fact, it has been speculated that not knowing the candidates personally (in a system that requires voters to specify a preferred candidate) explains the considerably lower turnout levels among the young, especially in municipal elections with a larger number of candidates (Borg, 2018, 131).Thus, while a larger proportion of candidates get elected locally, young candidates who lack a personal support base and experience from local politics may be disadvantaged.
There are also several arguments in favor of young aspirants "flying over" the traditionally viewed pathways to higher office, especially in the Finnish OLPR system where parties are compelled to reserve some seats in their ballots to young candidates and hence specifically promote them for national elections.In parliamentary elections, Finland is geographically divided into 13 electoral districts 1 in which each party may nominate at most 14 or, if more than 14 representatives are elected from the district, at most the number of candidates elected (in the 2023 elections, the maximum was 37).The number of representatives is proportional to the number of Finnish citizens residing in the district six months prior to the elections. 2In comparison to municipal elections, in parliamentary elections parties overall nominate far fewer candidates, which also means that the supply of aspirants is not an issue, at least not for the largest parties.
Candidates in Finland are heavily dependent on the number of personal votes they receive, which incentivizes them to focus on personalized campaigns and highlighting their individual characteristics and achievements, rather than cultivating a party vote only (Arter and Söderlund, 2023;von Schoultz, 2018).Since all preference votes are pooled together, parties do not need to worry about vote distribution between candidates.Instead, they focus on ensuring that their list of candidates is diverse and caters to a wide range of voter groups, following a "strategy of balanced lists" (Arter, 2013, 104).Previous research indicates that Finnish parties prioritize gender, age, occupation, and place of residence when balancing their lists (Arter, 2013;Sipinen, 2021, 215-219).
The strategy of balanced lists is so important to Finnish parties that in most of them, membership ballots have become the exception rather than the rule when nominating candidates in parliamentary elections.This is because membership ballots are not likely to produce a winning list, where candidates from various social groups would be represented, since parties' own organizations and membership ballot participants no longer mirror their electorates and because party members might value party loyalty over other more prominent vote-earning attributes (Arter, 2021).
When considering the "value added" of young candidates to a party list, then, parties will likely consider youth voter turnout and the degree to which this voting bloc can be persuaded to vote for their party.Previous research, as well as statistics data from Finland demonstrate that young voters tend to participate more in high-visibility elections, or national elections than in sub-national elections (Fortin-Rittberger et al., 2019;Statistics Finland, 2019, 2021).In the 2019 parliamentary elections, where the overall voting percentage was 72.1, the turnout rate among 25-34-year-olds was 62.9 percent and among 20-24-year-olds 58.8 percent.In the 2021 municipal elections (overall voting percentage 55.1), the turnout rates among the corresponding age groups were, 40.2 and 34, respectively.Consequently, parties might prioritize young candidates at the local level to tap into the significant pool of young persons eligible to vote who might otherwise abstain from voting if there are not young candidates on their preferred party's list.

Data and methods
To test our hypotheses about young candidates' selection and election on the local and national level of politics, we draw on two main data-sources: Registry data of electoral candidates and representative election surveys of voters.
Our first analysis utilizes a comprehensive dataset that includes detailed register data provided by the Finnish Ministry of Justice (2023).
1 There were 15 electoral districts up until a reform in 2015, where the number was reduced to 13.
2 However, from the electoral district of Åland is always elected one representative.
The register contains information about all individual candidates who have run in Finnish national and municipal elections over the past decade.Thus, we have information about candidates' age, gender, party affiliation, incumbency, and electoral success, as well as turnout-levels in the respective (local/national) constituencies.Furthermore, we include information about the district magnitude (the number of seats candidates are competing for in the respective constituency) and about the share of incumbents running for election.Our dataset includes candidates running in the municipal elections of 2012, 2017, and 2021, as well as the national elections of 2011, 2015, and 2019, yielding a total sample size of 113,299 candidates.Table 2 provides an overview of our primary variables of interest.
In a subsequent analysis, we make use of two representative election surveys of Finnish voters relating to the local elections in 2017 (N = 1727) and to the national elections in 2019 (N = 1599).These surveys asked about how important candidate age was for voters' decision, how important the candidates' prior political experience was, and whether people voted for a candidate that had roughly the same age as themselves.We draw on these items to shed some light onto possible explanations regarding the different dynamics in local vs.national elections.
Our main objective in the analysis of the registry data is to examine the role of age in candidate selection and election, comparing candidates in national and local elections.Our primary variable of interest is thus candidate age, which we use both as a continuous variable and to compute an indicator of whether a candidate in the sample is "young".In the models presented in the main text, we define a young candidate as being under 30 years old, following the common recommendations from the youth representation literature (Sundström and Stockemer, 2021).We also provide robustness tests with cut-off ages of 25 and 35 in appendix A. As seen in Table 2, the mean age of candidates in the sample is 49 years; and candidates under 30 constitute about 9% of the sample.3Fig. 1 visualizes the distribution of candidates′ age across political levels.
The second dependent variable we analyze is electoral success, operationalized as a dichotomous variable that takes the value of 1 if a candidate has been elected and 0 if not.Unsurprisingly, the majority of candidates are not successful in their election, with three fourths of them running without being elected.The main independent variable we are testing is an indicator variable for the type of election: national or municipal.As can be seen in Table 2, candidates in municipal elections are highly overrepresented and constitute most observations in the dataset.
As control variables, we include the party of the candidates as fixed effects in all models.Previous research has shown that party affiliation is the most significant determinant of both candidate selection and election (Cheibub and Sin, 2020;Schwindt-Bayer et al., 2010), and we can also assume that parties differ in the extent to which they promote young candidates (see Belschner and Dobbs in this issue).Additionally, we include a dummy variable for candidate gender, as young candidates may be more likely to be women (Belschner, 2023;Stockemer and Sundström, 2018), which could potentially confound the results.All models furthermore include a linear term for the year of election to control for possible time trends as well as district-level controls for the number of seats and the share of incumbents running.The models estimating the probability of electoral success additionally control for turnout levels.
Finally, in some models, we include information about the incumbency of the candidates.The data contain two incumbency variables, one for a sitting local councillor (inc.mun) and one for a sitting politician in the national parliament (inc.mp).As incumbents tend to be older and have a significant advantage in terms of electoral success (Ansolabehere et al., 2000;Ashworth and Bueno de Mesquita, 2008), some age-related disadvantages may in fact be explained by a lack of incumbency. 4 To control for both effects related to possessing a specific seat and having more general experience and name recognition, we include both incumbency terms as well as the respective interactions with the type of election in the models with incumbency controls (see Eq. ( 2)).Importantly, we do not control for a candidate's list position.In the Finnish electoral system, candidates do not appear on the ballot in an order pre-determined by the party (a system in which higher list spots would indicate a preference of the parties for the respective candidates).In contrast, candidates appear on the ballot in an alphabetical order.Additionally, while all lists are ordered alphabetically, they do not always start with the letter A, but randomly vary the letter the lists start with.
We employ linear probability (OLS) models to estimate how the political level of an election affects the age of selected candidates and the relative impact of age and being young on electoral performance.As the outcome variables we are estimating are partly dichotomous, we estimate all models with heteroskedasticity-robust standard errors. 5Additionally, we cluster standard errors on the electoral district level to account for the fact that candidates within districts may not be independent observations. 6The empirical models for the respective dependent variables take the following form:

Results
To test hypotheses 1 and 2 on how the political level affects candidate selection, we investigate how candidate age varies between local and national elections.As shown in the first column of Table 3 and in Fig. 2, candidates running in national elections are, on average, four years younger than those running in local elections.The mean age of candidates in municipal elections is 52 years, while in national elections it is 48 years.The second model, which uses the "young" dummy variable (with a cut-off age of 30) as its outcome, confirms these results.On average, the probability that a candidate is under 30 years old is 8.5% for local elections and 14% for national elections (Fig. 3).The control variables work as expected, showing that women tend to be younger, whereas incumbent candidates tend to be older.These results provide support for the "parachute" hypothesis (H2), which posits that candidates running in national elections are comparatively more likely to be young than those running in local elections.The results are similar in size and significance when we adjust the cut-off age to consider a candidate "young" downwards to 25 years or upwards to 35 years (Tables A.1 and A.2 in the Appendix).
Next, we investigate whether being young has a different effect on electoral success in national and local elections.To do this, we interact the election dummy variable with our "young" indicator variable.The first model for this dependent variable (model 3) is estimated without individual incumbency controls and thus shows the net difference between young and old candidates.It is not surprising that it is generally easier for candidates to succeed in municipal than in national elections.However, contrary to what was suggested in the "stepping stone" hypothesis, young candidates are significantly disadvantaged compared to older candidates in local elections, but not in national elections (see model 3 and Fig. 3).
However, to what extent are these results driven by the age-sensitive distribution of incumbency and political experience in general?To understand these effects, we added a full set of individual incumbency controls to model ( 4).The coefficients for model (4) in Table 3 suggest indeed that much of the age-specific disadvantage faced by young candidates on the municipal level is driven by incumbency disadvantages or, in other words, by a lack of political seniority.In model ( 4), the difference between the levels regarding the relative impact of age on electoral success becomes insignificant.The plot in Fig. 4 (based on estimates from model 4 including incumbency controls) illustrates these findings: While the absolute likelihood of success is higher in municipal elections, there are virtually no differences between young and older candidates once accounting for incumbency disadvantages.Tables A.1 and A.2 replicate these findings for cut-off-ages 25 and 35 and, while yielding different effect sizes, confirm that the main source of agerelated disadvantage on both levels is related to young candidates' lower likelihood to run as incumbents. 5Since the data we are dealing with is clustered on the party-and districtlevels, linear probability models with heteroskedasticity-robust and clustered standard errors have been shown to perform better and produce less biased results than logistic regression models (Gomila 2021;Huang 2016Huang , 2023;;Wooldridge 2002, 454). 6While we contend that the clustering of the data could adequately be modelled in a multi-level framework, we deal with a low number of highestlevel clusters (election).Therefore, multilevel models run into convergence issues and can produce biased and/or unstable estimates (Hox et al. 2017).We invite future research to use cross-national data with a larger number of elections to replicate our analyses in a multi-level framework.
In sum, these results support a parachuting effect at the candidate selection stage.As we have argued theoretically, this is likely an effect of higher candidate supply.First, candidate supply may be comparatively higher for national elections, as national politics tends to be a more attractive place to start a political career.It might also be the case that parties are more engaged in recruiting young promising candidates when composing lists for national elections.
At the same time, regarding election dynamics, the results neither support the stepping stone nor the parachute hypothesis.While the age disadvantage appears slightly larger in municipal elections, this is fully explained by incumbency disadvantages.As soon as incumbency and its correlation with age and political seniority is considered, there is no significant difference between the levels regarding young candidates' likelihood of winning an election.The non-existence of an agedisadvantage in national elections could partially be explained by the dynamics of party demand.Parties may only select the most promising young candidates for national electoral contests, while being less selective in local elections.At the same time, given the great weight of voter preferences in the Finnish system, we need to consider voter demand as a further explanatory factor.To understand dynamics of voter demand better, we conducted an additional exploratory analysis based on election survey data.
Fig. 5 displays findings from representative surveys of voters in the latest Finnish Municipal and National Election Study, where respondents were asked three separate questions regarding candidate age and their voting decision.The first two panels display the share of respondents who said that age (left panel) and prior political experience (middle panel) of candidates were an important or very important factor in their voting decision.First, on average, prior political experience is more important than age in both municipal and national elections.However, there is substantial heterogeneity across age groups.For young voters up until 29 years, candidate age was a more important characteristic than experience in municipal elections; suggesting that they might have voted specifically for young candidates.This is further supported by the graph in the right panel; indicating that more than 60% of respondents under 30 voted for a candidate of the same age in the 2017 local elections.In the national elections two years later, only about 35% of the young voted for a same-age candidate.Affinity, i.e., the tendency to vote for a candidate of the same age, is even higher among middle-aged voters.Approximately half of those aged between 30 and 59 voted for a candidate of their age in the national elections; so did more than 60% in local elections.In contrast, affinity is lowest for senior citizens over 60 years.
We additionally regressed a respondent's likelihood to indicate that candidate age/experience was important and their likelihood of same    *** p < 0.001; ** p < 0.01; * p < 0.05.Note: Linear probability models.Parentheses show heteroskedasticity-robust standard errors, clustered on the district-level.Models with incumbency controls include terms for both municipal and national incumbency as well as two interaction terms with the type of election.District controls include district magnitude and the share of incumbents of all candidates.See the full models with coefficients for all interactions in appendix B.   (Cutler, 2002).Importantly, the age groups do not constitute equal shares of the population.In the representative sample drawn from the Finnish voting population, young people only represent between 11 (municipal election) to 13 (national election) percent of all voters.In contrast, senior citizens alone make up 41% of the voters in municipal and 39% of voters in national elections (see Fig . C.2 in the appendix).Thus, the age preferences and affinities of older voters will weigh much more heavily in influencing election results.To illustrate how the relative weight of age groups and their preferences can influence the election results, consider a hypothetical example of a district with 20 seats.In municipal elections, approximately 2 of the candidates (8%) would be young, and 6 of them (29%) would be seniors, i.e., over 60 years old, according to the models in our first analysis.Assuming that, as indicated in Fig. 6, about half of the young voters (who represent 11 percent of the total electorate) would support a young candidate, whereas 30% of the seniors (who represent 41% of the electorate) would support a senior candidate, 1 of 20 seats in a local election would be allocated a young candidate due to affinity voting, whereas affinity effects would allocate 2.5 seats to senior candidates.In contrast, in national elections, 3 out of 20 candidates would be young, and 4 would be senior.When we again approximate the share of young and senior voters who vote for a same age candidate by the numbers in Fig. 6 (30% of young and 10% of seniors vote for a same age candidate) and consider the different shares of the age groups in the electorate, young and senior candidates would both be allocated 0.8 seats of 20 due to affinity voting. 7hese analyses provide three interesting additions to our previous results.First, young voters do seem to actively support young candidates, and they do so specifically in local elections.Second, the older a voter is, the more emphasis they put on the prior political experience of candidates in their voting decision.Also, senior citizens were particularly likely to vote for a same-aged candidate in municipal elections and least likelycompared to other age groupsto support a same-age candidate in national elections, which at least partly relates to lower supply of same-aged candidates at this level.Third, senior citizens constitute by far the largest age group in the Finnish electorate.Thus, their candidate age preferences weigh more heavily than those of young citizens, which affects the composition of candidate lists and elected bodies.

Discussion and conclusion
Our findings from Finnish national and local elections show, first, that young people are under-represented as candidates at both levels compared to their share in the population.However, in contrast to traditional expectations of hierarchical political stratification, where political aspirants are expected to start at the "lowest rung" of politics and work their way up the so-called political career ladder, our results suggest that, in some electoral systems and contexts, young aspirants may "fly over/parachute" the traditionally viewed pathways or stepping stones to higher office.Indeed, our findings show that candidates running in national elections are comparatively more likely to be young than those running in local elections.This is, arguably, mostly an effect of differences in candidate supply.While it is rational from parties′ point of view to diversify their lists in order to attract preference votes from young voters in elections on both political levels, the supply of young candidates appears to be comparatively larger in national elections: There are less spots that need to be filled and national politics should generally be more attractive to start a political career.
Secondly, we show that although the absolute likelihood of success is higher in municipal elections, the initial performance gap between young and older candidates is larger in local than in national elections.This effect does however disappear as soon as we control for incumbency, suggesting that incumbency disadvantages, that are highly correlated with age and political seniority/experience, explain any youth disadvantage in elections.At the same time, we also demonstrated how the age composition of electorates can affect election results.As senior voters were more likely to vote for same-age candidates in municipal than in national elections, while at the same time being much more numerous than young voters, their preferences would lead to a larger electoral advantage for senior candidates in local elections.
Our findings were generated by exploiting data from Finland.The uniformity of the Finnish electoral system across local and national elections enabled us to directly compare the effects of age in electoral contests at both levels.However, it is challenging to speculate on whether we would observe similar 'stepping stone' or 'parachute' dynamics in other electoral systems.As we have argued, such outcomes would depend on the specific configurations of supply and demand logics in conjunction with their respective local and national electoral systems.For example, our findings might differ in a system where preferential voting is employed only in local elections, while national elections use closed lists.Therefore, we encourage further research to systematically compare age effects in elections across a broader range of electoral and party systems.
One drawback regarding youth representation in general is, of course, the low average success rate in national elections.Our findings therefore suggest that in order to improve young people's political representation effectively, parties must place greater emphasis on the nomination of young candidates, particularly at the local level.This may require parties to rethink their recruitment networks and strategies, and to consider new ways of engaging with young people in order to attract them to party politics.We have also shown that voters use age as a heuristic cue, and that young voters are more likely to vote for young candidates.Therefore, parties that nominate more young candidates may also see an increase in voter turnout among young people.
We do not aim to exhaust all explanations for variations in candidacy and electoral success of young people, rather we aim to provide another piece of the puzzle around youth's overall performance in electoral politics.As such, while we do not test the full range of potential mechanisms that may explain variations in youth's representation as candidates and elected officials at the sub-national and national levels, we do believe that future studies should take an interest in youth's representation at different levels and should seek to investigate a range of mechanisms that may explain these variations.For instance, the exact way how incumbency theoretically and empirically conditions age effects, as well as how the contextual conditions (district-level characteristics, party affiliation, urban/rural divides, etc.) affect young candidates' chances to be selected and elected deserves further attention.

D Hypothetical Election Scenario
Hypothetical Scenario with 20 Candidates Municipal Elections: Composition of Electorate and Candidate list: • Young candidates (8% of 20): 1.6 (rounded to 2 for practicality) • Senior candidates (29% of 20): 5.8 (rounded to 6 for practicality) • Young voters (11% of all voters) who vote for young candidates: 50% of 11% = 5.5% of all voters • Senior citizens (41% of all voters) who vote for senior candidates: 30% of 41% = 12.3% of all voters Voting Scenario: • Young candidates would secure seats corresponding to 5.5% of all voters, which equals ap-proximately 1.1 seats.
• Senior candidates would secure seats corresponding to 12.3% of all voters, which equals approximately 2.46 seats.
• Senior candidates would secure seats corresponding to 3.9% of all voters, which equals ap-proximately 0.78 seats.

Fig. 1 .
Fig. 1.Candidate age distribution across elections.Note: plot shows the distribution of candidates′ age across all local and national elections included in the dataset.N = 113,299.

Fig. 2 .
Fig. 2. Predicted probability of candidate being young (m2).Note: Plot shows the predicted probability of candidate being younger than 30 years of age in municipal vs. national elections.Predictions are based on model 2 in Table3.Vertical bars show 95% confidence intervals.

Fig. 3 .
Fig. 3. Predicted probability of electoral success (without incumbency controls) (m3).Note: Plots show the predicted probability of a candidate being elected in municipal vs. national elections.Predictions are based on model 3 and 4 in Table3.Vertical bars show 95% confidence intervals.

Fig. 5 .Fig. 6 .
Fig. 5. Voter preferences by age group.Note: plots show share of people who agree completely or a lot with the following statements.Left panel: "To what extent did the candidate's age influence your choice of candidate?",middle panel: "To what extent did the candidate's prior experience in politics influence your choice of candidate?",right panel: "What was the age of the candidate you voted for?".Colours distinguish respondents by age group.Sources: Finnish Municipal Election Study 2017 (n = 1465) (Borg, 2017) and Finnish National Election Study 2019 (n = 1315) (Grönlund and Borg, 2019).(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. C. 1 .
Fig. C.1.Shares of age groups in the electorate Note: Figure shows shares of age groups among voters in municipal elections 2017 and national elections 2019.Source: Finnish municipal/national election study.
H3: Young candidates are more successful in municipal elections than in national elections.H4: Young candidates are more successful in national elections than in municipal elections.

Table 2
Summary statistics.

Table 3
Multivariate regression results.
Table B.1.Sources: Finnish Municipal Election Study 2017 (n = 1465) and Finnish National Election Study 2019 (n = 1315).agevotingontheiragegroup and the respective survey (national vs. municipal).The results are shown in TableC.1 (Appendix) and in Fig.6.First, candidate age matters most for young respondents, and mostly so in municipal elections.Interestingly, while candidate age is decidedly more important in municipal than in national elections for young respondents (up until age 39), the difference becomes insignificant for older respondents.As the middle panel displays, the importance that voters place on previous experience almost linearly increases with age, but only the oldest respondents place significantly more emphasis on experience in national than in local elections.Third, almost all groups (with the exception of the middle-aged respondents) had a higher likelihood to vote for a candidate roughly the same age in local elections than in national elections.This gap is particularly large for senior citizens: They had an average probability of about 30% to vote for a same age candidate in local elections, compared to only 10% in national elections.The generally higher affinity effects in municipal vs. national elections are consistent with previous research suggesting that affinity is most pronounced in elections where voters have little information and knowledge about the specific candidates **p < 0.01; **p < 0.05; *p < 0.1.Note: Linear probability models.Parentheses show heteroskedasticity-robust standard errors.Reference categories for factor variables: 40 to 49; municipal survey.Survey weights are included.