Too Young to Run? Voter Evaluations of the Age of Candidates

Charles T. McClean∗1 and Yoshikuni Ono†2

1Harvard University 2Waseda University and the Research Institute of Economy, Trade and Industry

June 29, 2021‡

Abstract In most countries, elected officials tend to be older than most of their constituents. Is this because voters generally prefer older politicians to younger ones? We inves- tigate this question by conducting two novel survey experiments in Japan. We ask respondents to evaluate the photos of hypothetical candidates while altering candi- dates’ faces using age regression and progression software to make them appear as if they are younger or older. Contrary to the observed age demographics of politicians, we find that voters dislike older candidates the most, but view younger and middle- aged candidates as equally favorable. Moreover, while younger and middle-aged voters prefer candidates closer to their own age, older voters do not. Voters rate older can- didates as the least competent, least likely to focus on most policy issues, and least electable. Together, these findings suggest that supply-side factors rather than voter demand explain the shortage of younger politicians.

∗Postdoctoral Fellow, Program on U.S.-Japan Relations, Weatherhead Center for International Affairs, Harvard University, 61 Kirkland Street, Cambridge, MA 02138. Contact: [email protected]. †Professor, Faculty of Political Science and Economics, Waseda University and Faculty Fellow, Research Institute of Economy, Trade and Industry. 1-6-1 Nishiwaseda, Shinjuku, Tokyo 169-8050 Japan. Contact: [email protected]. ‡The authors are grateful for the financial support provided by the Research Institute of Economy, Trade and Industry (RIETI), Frontier Research Institute for Interdisciplinary Sciences (FRIS) at Tohoku University, Waseda University, the Okawa Foundation for Information and Telecommunications, and the Japan Society for the Promotion of Science (Grants-in-Aid for Scientific Research 20H00059; 19H00584; 19H01449; 18H00813). This study utilizes the micro-level data of the questionnaire from the “Survey on Attitudes Toward Politics, Society, and the Economy,” which was conducted by RIETI. The authors would like to thank Andr´eBlais, Daniel Butler, Charles Crabtree, Gordon Dahl, Shusei Eshima, Yusaku Horiuchi, Peter John Loewen, Megumi Naoi, Stefanie Reher, Margaret Roberts, Kaare Strøm, Daniel Smith, Hikaru Yamagishi, and participants in the sessions at the Asian Politics Online Seminar Series and the 2020 American Political Science Association Annual Meeting for their helpful comments. Politicians vary significantly in age, but in most countries, elected officials tend to be older than most of the constituents they represent. In the United States, the 2020 presidential race pitted Joe Biden (77) against Donald Trump (74), the oldest candidates to ever stand for the office. U.S. millennials (born between 1981 and 1996) are now the largest living generation and largest share (31%) of that country’s electorate, yet they occupy just 26 (6%) of the 435 seats in the House of Representatives. They are outnumbered 9 to 1 by baby boomers (born between 1946 and 1964), who hold 233 seats.1 In Japan, the focus of this study, young people are similarly underrepresented in public office. Nearly a third of the voting-age population is under age 40, but fewer than 10% of elected positions at any level of government are held by citizens under 40.2 Recently, scholars have begun to pay more attention to the age bias inherent to political institutions. One body of research discusses how the underrepresentation of young people raises normative concerns about intergenerational equity (e.g., Bidadanure 2015). Others provide evidence that just as a shortage of women or racial and ethnic minorities in elected office can affect policy outcomes on issues related to gender and race (e.g., Chattopadhyay and Duflo 2004; Franck and Rainer 2012), a relative lack of younger politicians has demon- strable consequences for age-related government spending and social welfare programs (Curry and Haydon 2018; McClean 2021). Some evidence even suggests that a de facto gerontoc- racy could be self-fulfilling, in that it discourages young people from participating in elections (Pomante and Schraufnagel 2015). Among these nascent studies, most have focused on seeking out supply-side explanations for the emergence of government by older people. In the U.S. context, two recent books sug- gest that young people have less ambition to become politicians because they feel alienated from contemporary politics; view elected officials as corrupt, dishonest, and inefficient; and believe that they can best enact change in their communities through other means (Lawless

1Data is for the 116th Congress (2019–2020). Drew Desilver, “Millennials, Gen X Increase Their Ranks in the House, Especially Among Democrats,” Pew Research Center, November 31, 2018. 2Statistics Bureau of Japan (2004–2019); McClean (2021).

1 and Fox 2015; Shames 2017). A handful of cross-national studies, on the other hand, point to political institutions such as candidate-centered electoral systems and minimum age re- quirements for specific elected offices as the main culprits behind graying government (Joshi 2013; Stockemer and Sundstr¨om2018). In this article, we test a separate, demand-side explanation: that voters prefer older politicians over younger ones. That is, young people may be underrepresented in office because voters hold an overall age bias against younger candidates (youth discrimination hy- pothesis). Another, not mutually exclusive explanation is in-group favoritism, where voters prefer candidates closer to them in age, but young people turn out to vote at lower rates than older people, resulting in an overrepresentation of older elected officials (in-group fa- voritism hypothesis). In addition to testing these two explanations, we investigate potential mechanisms behind them by analyzing whether voters form candidate preference heuristics by inferring certain characteristics about politicians based on their ages. For instance, voters may build from stereotypes to assume the policy issues and governance styles an older or younger candidate would pursue if elected. Furthermore, we consider voters’ perceptions of electability, which may also vary with the age of candidates. To test for age biases and stereotypes among voters, we conduct two novel survey exper- iments in Japan. We ask respondents to evaluate the photos of hypothetical candidates for mayor, which we have manipulated using age regression and progression software to make the candidates appear as if they are younger or older. By taking advantage of recent advances in machine learning and neural networks, we are able to hold constant those elements of each photo unrelated to aging—including the candidate’s expression, underlying facial structure, and clothing—and manipulate only those elements that tend to change with age. In the first experiment, we test the demand-side explanation—age biases—by showing respondents two candidate photos with randomly manipulated ages, then asking for whom they would vote for if these two candidates appeared on the ballot. In the second, we explore the possible mechanisms behind the demand-side explanation—age stereotypes—by

2 presenting a separate group of respondents with individual candidate photos, asking them to assess each candidate’s likely issue emphases, traits, and electability. Contrary to the observed underrepresentation of young people in elected bodies, we find that voters dislike older candidates the most. Respondents were significantly less willing to vote for older candidates, but were equally willing to support younger and middle-aged candidates. This elderly bias was also reflected in our tests for in-group favoritism. While younger and middle-aged voters were more favorable toward candidates from their age group than others, older voters were, if anything, more critical of elderly candidates than others. Why do voters dislike elderly candidates? In our second experiment, we find evidence that voters hold significant stereotypes about age. Respondents drew clear links between a candidate’s age and their issue emphases, traits, and electability. Younger candidates were thought to focus more on education, childcare, climate change, anti-corruption measures, and foreign residents; middle-aged candidates on the economy and safety; and older candidates on elderly care and healthcare. Middle-aged candidates were seen as the most electable and having the most favorable traits for office. Older candidates were panned—viewed as the least competent, least likely to focus on most policy issues, and least electable. In sum, we find little evidence that voter biases lead to youth underrepresentation in office. Voters in our experiment were equally willing to support younger and middle-aged candidates, and they preferred younger candidates substantially over older candidates. Our findings thus indicate that it is likely supply-side factors, such as political ambition and institutional barriers, that most inhibit young people from holding office. If institutions are reformed and young people are encouraged to run, our results suggest that voters may welcome their greater presence in public office.

Voter Evaluations of the Age of Candidates

When evaluating candidates, voters make inferences about their likely policies and effective- ness in office based on a wide range of heuristics, from those more directly connected to

3 politics, such as a candidate’s party and endorsements (Rahn 1993; Sniderman, Brody and Tetlock 1991), to characteristics such as a politician’s gender, race, class, and even physical appearance (Aguilar, Cunow and Desposato 2015; Burden, Ono and Yamada 2017; Carnes and Lupu 2016; Todorov et al. 2005). However, despite significant public attention to the age bias of political institutions,3 we know little about whether voters infer candidate infor- mation based on age and, if so, how it affects their decision-making at the ballot box. If voters dislike younger politicians, for example, then young people may choose not to run for office as often as older people, or they may run but simply lose much more often at the polls. Reviewing the extensive research on voter biases concerning gender, race, and ethnicity, it is easy to imagine that voters might similarly discriminate against candidates based on their age. Research shows, for instance, that voters form stereotypes about female and racial or ethnic minority candidates (Kage, Rosenbluth and Tanaka 2019; Kam and Kinder 2012). For example, voters often view female candidates as more liberal, compassionate, and competent on certain issues such as gender equality, healthcare, and education, but less assertive or qualified to handle security and economic policies than men (Huddy and Terkildsen 1993; Lawless 2015). There is also strong evidence from a wide range of contexts that voters are more likely to support coethnic candidates (Adida 2015; Shockley and Gengler 2020). These biases can help explain why so few women and minorities hold office (Dolan 2004; Ono and Yamada 2020), although they may play a less significant role in settings in which other cues, such as party affiliation, are especially influential (Aguilar, Cunow and Desposato 2015; Lawless and Fox 2015). If anything, voters may feel relatively more comfortable acting on age stereotypes because in many contexts such biases are viewed as a more acceptable form of discrimination. While there are strong social norms that a person’s gender or race should not exclude them from office, age is closely connected to experience, energy, health, and mental acuity, all of which

3See for example the UN’s Not Too Young to Run campaign or media coverage such as Derek Thompson, “Why Do Such Elderly People Run America?” The Atlantic, Mar. 5, 2020, and Isabel Reynolds and Emi Urabe, “Japan Is Too Old-Fashioned, Says One of Youngest Ministers Ever,” Bloomberg, Sept. 11, 2019.

4 are seen as relevant to judging a person’s ability to serve as an elected representative. Ageism against younger candidates also differs from other types of discrimination in that youth is both universally experienced and temporary: voters may feel it is alright to discriminate against younger candidates because these candidates will eventually become older and have their time to serve in office.4 Moreover, while many countries have sought over time to remove restrictions to encourage greater participation by women and racial or ethnic minorities in elected office, codified discrimination against younger citizens has largely remained in the form of legal rules that set the minimum ages for a person to be eligible to vote or stand for office.5 These restrictions are similarly seen as more understandable in light of concerns about young people’s cognitive development, experience, and maturity, even though the age minimums can be as high as 40 years old for legislative offices and 50 for executive office in some countries.6

Age Biases

Our study focuses on two sources of age bias that may contribute to the shortage of younger politicians: youth discrimination and in-group favoritism. By youth discrimination, we mean the electorate’s shared beliefs about the relative ability of younger people to serve in public office, and by in-group favoritism, we refer to voter preferences for candidates closer to their own age. The former could lead to a shortage if voters share a widespread dislike for younger politicians, while the latter would function via voter turnout—older voters turn out to vote at higher rates than younger voters, therefore in-group favoritism would suggest better ballot box odds for older candidates. In terms of youth discrimination, there are good reasons to believe voters may view certain candidates as “too young to run.” These beliefs may be rooted in either political

4Age discrimination also does not come with the same level of historical exclusion as discrimination based on other ascriptive characteristics such as gender and race (Mansbridge 1999; Phillips 1995). 5Some countries have recently lowered the age of candidacy to match the minimum age requirement for voting, such as the United Kingdom, which lowered it from 21 to 18 in 2007. 6The minimum age of candidacy is 40 for the upper houses of Cameroon, Czech Republic, Rwanda, and Zimbabwe, and one must be at least 50 to be president of Italy.

5 or non-political factors. Regardless of a candidate’s political record, some voters may view candidates below a certain age as simply lacking the necessary life experience and maturity to capably represent them in office. As for more political reasons, prior research shows that voters are more likely to support candidates with greater name recognition, local ties to the community, financial resources, political networks, and experience, especially prior experience in elected office (Jacobson 1983; Shugart, Valdini and Suominen 2005; Tavits 2009). All of these factors can be helpful to a politician’s success in office, yet they are also ones that individuals tend to accrue with age.7 Voters may therefore discriminate against younger candidates because they infer that young people lack the political capital resources to be either competitive candidates or effective representatives. It is also possible that negative sentiments regarding younger voters could translate to a negative bias against younger candidates. In part because younger voters have lower turnout, the political arena is filled with critiques about the participation of younger generations in politics. As Holbein and Hillygus (2020, 7) write, young citizens have been described as “apathetic, disengaged, narcissistic, selfish, entitled, shallow, lazy, impulsive, confused, lost, impatient, and pampered.” The conventional wisdom is that young people generally lack the interest, sense of civic obligation, or skills to participate in the electoral process, although they are often quite active in political engagements other than voting (protests, for instance) (Dalton 2008; Holbein and Hillygus 2020; Wattenberg 2007). Voters may therefore draw on these familiar narratives about younger people’s apathy toward electoral politics when considering younger candidates and view them in a more negative light. Alternatively, voter discrimination against younger candidates may be driven less by a relative animosity toward young people and more by a shared belief that the “optimal” age for a politician is significantly older. The large literature on age discrimination in the workplace, for example, yields several examples in which workers believe there are “correct ages” for certain occupations (Posthuma and Campion 2009). Retail, sales, technology, and

7This is one reason why younger politicians are often more likely to be part of political dynasties, as family support enables them to accumulate these important resources at an earlier age (Smith 2018).

6 finance are often seen as particularly young industries (McGoldrick and Arrowsmith 2001), whereas jobs that require more managerial skills are typically associated with older workers (Cleveland and Hollmann 1990). Negative stereotypes are often strongest when workers do not match the perceived correct age for the position (Perry, Kulik and Bourhis 1996). If voters believe elected representatives similarly need managerial skills to be effective—as in, perhaps, trustee models of representation (Fox and Shotts 2009)—then voters may view older candidates as better suited to hold public office and judge candidates who are “too young” more harshly. While we are unaware of any studies that focus on voter preferences toward younger candidates in particular, a few studies have used conjoint experiments to more generally measure voter preferences regarding a candidate’s age. These candidate choice experiments only test for age discrimination, not in-group favoritism or age stereotypes, in part because age is typically included only as a control variable, and the findings are mixed and sometimes conflicting. Depending on the setting and the inclusion of other variables, these studies alter- natively find that voters prefer middle-aged candidates over younger and elderly candidates, younger candidates over middle-aged and elderly candidates, or have no preference (Arne- sen, Duell and Johannesson 2019; Clayton et al. 2019; Eshima and Smith 2021; Horiuchi, Smith and Yamamoto 2020; Kirkland and Coppock 2017; Ono and Burden 2019). While the evidence of youth discrimination is mixed, most studies find evidence of two, somewhat conflicting patterns: voters generally prefer candidates with more experience in elected office but also dislike candidates past a certain age. We focus in particular on age in our experiments to investigate whether youth underrep- resentation could be a result of voter discrimination against younger candidates. Our youth discrimination hypothesis can be stated as follows:

H1: Voters will be less likely to vote for younger candidates than older candi- dates.

Another, not mutually exclusive, demand-side explanation for youth underrepresentation

7 could stem from in-group favoritism. A large body of literature shows that descriptive representation matters for voters, with citizens tending to prefer candidates that share similar demographic characteristics (e.g., Mansbridge 1999; Phillips 1995). While most work on affinity voting has focused on gender, race, and ethnicity, however, our hypothesis is that similar effects may be visible with regard to age. Moreover, if this age affinity voting is present and substantial, then the shortage of younger elected officials could be a consequence of low youth turnout. In other words, since middle-aged and older people turn out to vote at higher rates than younger people, the age distribution of politicians would then reflect the age distribution of actual voters, rather than the distribution of the electorate. Why should people vote for a candidate closer in age? Perhaps the most significant reason is that descriptive representation can lead to substantive representation. That is, voters may prefer candidates closer to them in age because they expect these candidates will be more likely to act in their interests in office. Like women and racial or ethnic minorities, research shows that younger people have distinct policy preferences from older people on a wide range of issues including education, childcare, unemployment, healthcare, immigration, gender equality, global governance, and environmental protection (Kissau, Lutz and Rosset 2012; Norris and Inglehart 2001; Wattenberg 2007). Moreover, there is some evidence that elected officials are more likely to promote and implement policies important to voters of a similar age. For example, Curry and Haydon (2018) show that older members of the U.S. Congress are more likely to introduce legislation on senior issues such as elder abuse and late-life housing, while McClean (2021) finds that younger Japanese mayors are more likely to increase municipal spending on child welfare for younger families. These studies provide good reasons as to why voters may prefer similarly-aged candidates. Two key recent observational studies address age and in-group favoritism.8 In the first, Webster and Pierce (2019), using survey data from the Cooperative Congressional Election

8A third, related study is Pomante and Schraufnagel (2015), whose voter turnout experiment involved randomly presenting American college students with photos of candidates of different ages. The authors do not consider vote choice, but their results show that younger people are more likely to say they will participate in elections when they are presented with younger slates of candidates.

8 Study (CCES), report some support for the use of age-based heuristics in U.S. elections, most prevalently in low-information elections, among higher-educated individuals, and in the comparative judgment of co-partisan candidates. In the second study, Sevi (2021) tests voter preference for party leaders and presidential candidates closer to their own age, lever- aging cross-national data on 51 countries from the Comparative Study of Electoral Systems (CSES). She, too, finds evidence of age affinity in voting behavior, although she cautions that the effect sizes in her study are small and call for further study. Both of these studies thus show a correlation between a voter’s age and their preferred candidate’s age.9 We explore whether this age affinity voting is observable in an experimental setting and, if so, whether in-group favoritism combined with low youth turnout could potentially contribute to the underrepresentation of young people in elected bodies.

H2: Voters will be more likely to vote for candidates closer to them in age.

Age Stereotypes

To explore potential mechanisms behind youth discrimination and in-group favoritism, we further examine three types of stereotypes through which a candidate’s age could affect vote choice: issue emphases, traits, and electability. First, as discussed in the previous section, voters may infer that younger and older can- didates will be more likely to emphasize policy issues that are especially salient for similarly- aged voters in office. For example, voters may expect that younger candidates will be rela- tively more likely to focus on issues that disproportionately affect young people, from child- care and education to longer-term challenges such as climate change (Busemeyer, Goerres and Weschle 2009). By contrast, voters may associate older candidates more with pensions, senior services, and healthcare, as these policies especially affect elderly retirees (Goerres 2009). If these issues are important to voters when selecting candidates, as suggested by

9These studies are not without their limitations. Relying on observational data has the benefit of realism but raises concerns about selection bias. It might be that candidates from certain age groups share other characteristics apart from age that make them more likely to receive support from similarly aged individuals.

9 public opinion polling,10 then these age stereotypes may help drive in-group favoritism in particular. Thus, we hypothesize that:

H3: Voters will expect candidates to focus more on issues important to similarly- aged voters.

Second, as noted earlier, youth discrimination among voters may be rooted in a belief that young people lack the necessary traits to be successful candidates or elected officials. Voters may therefore dislike younger candidates because they view these candidates as too inexperienced, incompetent, or unreliable relative to their older competitors. Voters may also infer that inexperience will lead younger candidates to be less effective representatives, whether it be in assertively representing the interests of their constituents in office, or working with other political actors to build consensus and enact their agenda. If members of the electorate share a general agreement that young people lack these relevant traits, then this could be a mechanism underlying voter discrimination against younger candidates.

H4: Voters will expect younger candidates to have less favorable traits than older candidates.

Finally, voters may use a candidate’s age to make inferences about their likelihood of winning an election. There is a large body of research that shows that voters often vote strategically rather than sincerely (e.g., Duverger 1954). In many cases, voters may vote for a candidate they believe has a better chance of winning over the candidate they prefer. This strategic behavior could arise because voters do not wish to “waste” their vote if their preferred candidate has little to no chance of winning the election, or because voters receive some psychological benefit from joining the “bandwagon” and voting for a more popular candidate (e.g., Barnfield 2020). Thus, even if some voters like younger candidates, they may strategically choose to support older candidates because they view these candidates as

10See, for example, the University of Tokyo and Asahi Shimbun’s (UTAS) joint surveys of voters, available at http://www.masaki.j.u-tokyo.ac.jp/utas/utasindex.html.

10 more electable. Concerns about the electability of younger candidates could thus bolster youth discrimination among voters.

H5: Voters will expect younger candidates to be less likely to win an election than older candidates.

Age Regression and Progression Experiments in Japan

We test these hypotheses using an original experimental design in Japan where we manipulate the photos of hypothetical mayoral candidates using age regression and progression software. Japan provides an ideal setting to test for age discrimination against political candidates. First, young Japanese people have traditionally been underrepresented in their political institutions, yet the country is a well-known “graying” society (Muramatsu and Akiyama 2011). That is, age-related policy issues are salient, given that the country faces the twin demographic challenges of a declining birthrate and rapidly aging population (Kweon and Choi 2021; Umeda 2020). Japan is also a society of strong age norms, especially regarding respect for elders and clearly defined roles for younger and older citizens (Lee 2016). Second, Japanese campaigns are highly regulated and prominently feature the faces of candidates. We focus in particular on mayoral elections, in which nearly every candidate runs as an independent. Rather than relying on a party label, candidates must devote significant effort to developing a personal vote, often via copious posters plastered in high-traffic areas (Horiuchi 2005). Voters in Japan are thus accustomed to learning about candidates—even the fact that an election is occurring—via these enormous color photos of candidates’ names and faces (Lewis and Masshardt 2002). Furthermore, the candidate pool for mayoral races is largely homogeneous—over 99% are ethnically Japanese and 98% are men—making age one of the most distinct observable differences between candidates (McClean 2021). With that being said, younger mayors are less common in Japan. Japanese citizens must be at least 25 years old to run for mayor, but the median age of an elected mayor is 62 years old. If we divide the voting-age population into three, relatively equal groups (aged 18–44,

11 45–64, and 65 and over), we can see that younger voters are by far the most underrepresented in government. People under 45 make up 38.8% of the electorate, but just 4.4% of mayors. By contrast, middle-aged voters are clearly overrepresented: 31.4% of eligible voters are between the ages of 45 and 64, while 59.7% of Japan’s mayors fall into this age group. The final group, older voters, are also overrepresented, though to a lesser extent than middle-aged voters. Citizens aged 65 or older represent 29.7% of the voting-age population but 35.9% of mayors. The modal mayor in Japan begins their term at 65 years old—coincidentally also Japan’s national retirement age.11

Experiment 1: Age Biases

To test for age biases and stereotypes, we fielded two experiments in Japan. In Experiment 1, our goals were to test youth discrimination (in terms of the likelihood that a voter will support older candidates compared to younger candidates) and in-group favoritism (whether voters view candidates closer to their own age more favorably). Experiment 1 was administered in March 2020 by Rakuten Insight Inc., one of Japan’s major survey firms, to a randomly selected survey sample from Rakuten’s subject pool. This survey sample was matched to the population census in terms of respondent age, sex, and region of residence, and we aimed to collect a sample size of 3,000 citizens of eligible voting age. We received 2,901 valid responses.12 For Experiment 1, we licensed the photos of two different male Japanese models from Shutterstock, an image warehouse (Table 1). These photos are styled similarly to typical campaign materials: both men wear dark grey suits with brightly colored ties, smile slightly while facing the camera directly, and sport relatively conservative, common haircuts that could be seen on people of different ages. One man is even raising a clenched fist, using a

11Calculated using the Statistical Handbook of Japan (Ministry of Internal Affairs and Communications 2020) and the Japanese Municipal Elections Dataset (McClean 2021). 12The legal voting age is 18 in Japan. We collected 3,270 responses in total but removed 369 responses in the process of cleaning our data. First, we removed the 67 respondents who represented the slowest 1% and fastest 1% in completing our survey. Second, we removed 283 respondents who failed to answer the trap question to ensure respondents were paying attention to our survey. Third, we removed 19 respondents who claimed they could not see the candidate images.

12 Table 1: Manipulated Photos of Hypothetical Candidates for Mayor

Younger Middle-Aged Older

Model 1

Model 2

Notes: In Experiment 1, respondents were randomly shown two candidate photos at the same time, one of each model. In Experiment 2, respondents were randomly shown just one candidate photo. signal frequently adopted by candidates of all ages in Japanese elections. We used FaceApp, a mobile application created by Wireless Lab, in order to augment the pair of candidate photos to appear younger or older.13 Importantly, FaceApp’s age regression and progression algorithms only manipulate those elements of each photo that are likely to change with age, leaving the model’s underlying facial structure, expression, hairstyle, and clothes as well as the background of the original photo intact. Accounting for these factors is key, given the extensive literature on the ways different aspects of a candidate’s appearance, from attractiveness to smile, facial structure, and skin tone, can influence voter evaluations (Horiuchi, Komatsu and Nakaya 2012; Ono and Asano 2020; Terkildsen 1993; Todorov et al. 2005). FaceApp, in other words, allows us to sidestep a number of confounding factors that might otherwise affect our analysis. We created three versions of each model’s photo within the age range of usual mayoral candidates (see Table 1): one photo on the younger end of the spectrum, one middle-aged,

13Additional information on FaceApp can be found in the Appendix.

13 and one older. We asked respondents to guess each model’s age to verify that the photos met our expectations. Respondents estimated that the three versions of Model 1 were 29, 52, and 72 years old, and that the three photos of Model 2 were 30, 51, and 72 years old.14 In Experiment 1, we showed each respondent a pair of photos (one of each model) ran- domly selected from Table 1, then asked them to assume that these two people were candi- dates for a mayoral election. We mentioned that (i) neither candidate was the incumbent; (ii) both candidates were independents; and (iii) the election was in the municipality in which the respondent resided. The result is a 3×3 factorial design, in which the three levels of each treatment are the younger, middle-aged, and older versions of each model’s photo. Respondents were then asked which candidate they would vote for in the election.

Experiment 2: Age Stereotypes

To investigate the mechanisms underlying the age discrimination in Experiment 1, we fielded a second, parallel survey experiment. Experiment 2 was administered the same month (March 2020) as Experiment 1 as part of the “Survey on Attitudes Toward Politics, Society, and the Economy,” conducted by the Research Institute of Economy, Trade, and Industry (RIETI), a Japanese policy think-tank. RIETI used Rakuten Insight, the firm we used in Experiment 1, to field this survey. From Rakuten’s subject pool, we drew a representative sample of 3,000 new respondents of eligible voting age (matching the population census distribution in terms of respondent age, sex, and region of residence). No respondents overlapped Experiments 1 and 2. To test our three hypothesized age stereotypes, we showed respondents a single, randomly selected candidate photo from Table 1, then asked questions about the candidate’s likely issue or policy emphases, individual traits, and electability.15 In total, we asked about 11 different

14Respondents guessed the model’s age after our experimental questions. We also checked estimates for each model in a pilot survey (sample size: 300), which resulted in slightly older estimates. Respondents thought that the three photos of Model 1 were 36, 62, and 82 years old, and that the photos of Model 2 were 33, 58, and 80. 15One difference from Experiment 1 is that in Experiment 2, we told respondents that the candidate was running for mayor in a city with a population around 300,000. We specified the population of the city to avoid having respondents make their own inferences about the type of municipality and level of mayoral

14 policy issues and eight different traits.16 We selected the list of policies and traits based on our substantive knowledge about local government in Japan and by consulting past elite and public opinion surveys. In doing so, we sought to strike a balance between common issues and traits included in other studies of candidate evaluations as well as those we expected would have a particular connection with age (such as age-related welfare policies and prior political experience). For each of our mechanism questions, respondents answered on a 5-point Likert scale, ranging from “Very Unlikely” to “Very Likely” to emphasize the policy issue, exemplify the trait, or win an election. In the main text, we then dichotomize this scale by using the combined percentage of respondents who said “Likely” or “Very Likely” as our dependent variable. Each respondent completed the experiment twice, once for each model, with each experiment thus being a 1×3 design. To the best of our knowledge, our study is the first to use age regression and progression software in either a candidate choice (Experiment 1) or candidate evaluation (Experiment 2) experiment. Thanks to advances in machine learning and neural networks, applications such as FaceApp have grown increasingly sophisticated in recent years, making it easier for researchers outside of computer science to manipulate the perceived ages of photos. Studies in other fields, for example, have explored how exposing an individual to what they might look like in the future can affect their attitudes toward aging, the elderly, and harmful behaviors such as smoking (e.g., Flett et al. 2013; Rittenour and Cohen 2016). Age progression software is also commonly referenced in criminology research given its use by the police to help with the search for missing people and wanted fugitives (e.g., Lampinen et al. 2012). However, studies to date have yet to explore how voters react when presented with candidate photos

policy discretion. The Japanese central government can designate cities with a population over 200,000 as “core cities” (chukaku shi) and delegate to them certain functions and authority for issuing permits and licenses otherwise handled at the prefectural level. Local Autonomy Act, Article 252, Section 22. 16The 11 policy issues included education, childcare, climate change, anti-corruption measures, foreign residents and multiculturalism, economy and unemployment, public works, budget deficit, crime and safety, elderly care, and healthcare. The eight traits consisted of dominant, reliable, determined, considerate, competent, consensus-oriented, experienced, and long-term oriented.

15 of different perceived ages. The few candidate experiments that include at least a mention of age have typically relied on conjoint designs.17 Compared to our age regression and progression experiments, these analyses do have an advantage in their ability to control for a wide range of candidate char- acteristics and approximate an information-rich environment (Hainmueller, Hopkins and Yamamoto 2014). One disadvantage, however, is that evaluating long lists of candidate attributes against one another often differs from voters’ typical cognitive processes in eval- uating real-world candidates. Few voters collect detailed information on the entire slate of candidates before voting, relying instead on information shortcuts based on easy-to-observe characteristics of candidates such as their party, appearance, gender, race, and, per our contention, age. Our experiment’s comparative advantage is its focus on a low-information environment approximating the real-life process in which voters evaluate a candidate’s age via their ap- pearance, such as when they view a candidate poster in Japan. Studying the effects of the perceived ages of candidates in photos, as opposed to the chronological ages of candidates listed in conjoint experiments, is further important given that voters often infer a candidate’s age based on their facial appearance rather than learning their exact age via other means. Our design allows us to estimate age effects that would normally be very difficult to measure in an experimental setting: namely, how voters would evaluate a given candidate if that candidate somehow appeared younger or older.

Results

Age Biases

Do voters dislike younger candidates? We begin by considering what our age regression and progression experiments revealed regarding age discrimination among respondents. For ease

17One notable exception is Pomante and Schraufnagel (2015), but we improve on their research design by using manipulated photos of the same person rather than different models and photos, which vary along multiple dimensions beyond age.

16 Figure 1: Candidate Age and Vote Choice

(Baseline: Middle−Aged Candidate)

Younger Candidate ●

Older Candidate ●

−60 −40 −20 0 20 Percentage Point Change in Probability of Respondent Voting for Candidate

Notes: Bars show 95% confidence intervals.

of presentation, we average our results across the two models and collapse our 3×3 experiment into the three treatment conditions featuring hypothetical candidates who appeared different in age: Younger vs. Middle-Aged, Younger vs. Older, and Middle-Aged vs. Older. Figure 1 plots the difference in the probability that respondents said they would vote for the younger or older candidate, using the middle-aged candidate as the comparison condition.18 In sharp contrast to the idea that voters prefer older politicians over younger ones (H1), we find that voters actually dislike older candidates the most but are equally favorable toward younger and middle-aged candidates. When presented with a mayoral race between an older candidate and either a younger or middle-aged candidate, respondents were over 50 percentage points less likely to choose the older candidate. In contrast, respondents were slightly more likely to say that they would cast their vote for a younger candidate compared to a middle-aged candidate, although the difference is not statistically significant. Turning to our expectations concerning in-group favoritism (H2), Figure 2 breaks down our experimental results for vote choice into three respondent age groups: younger voters (18–44), middle-aged voters (45–64) and older voters (65 and over). As noted earlier, we begin with these age cutoffs because they divide the electorate into three, roughly equally sized groups that match up well with the intended ages of our candidate photos.

18In the Appendix, we provide the results disaggregated by model and show that they are robust to manipulation checks in Tables A1, A2, and A3.

17 Figure 2: Candidate Age and Vote Choice by Age of Respondent

Younger Voters Middle−Aged Voters Older Voters (18−44) (45−64) (65 and Over)

(Baseline: Middle−Aged Candidate)

Younger Candidate ● ● ●

Older Candidate ● ● ●

−60 −40 −20 0 20 −60 −40 −20 0 20 −60 −40 −20 0 20 Percentage Point Change in Probability of Respondent Voting for Candidate

Notes: Bars show 95% confidence intervals.

We find clear evidence of in-group favoritism among younger voters, but the evidence is mixed among middle-aged voters and absent among elderly voters. Younger voters pre- fer younger-looking candidates to middle-aged ones by 8.9 percentage points and to older candidates by a rather eye-catching 54.0 percentage points. Middle-aged voters had the most favorable opinions of middle-aged candidates, choosing them over older candidates by 59.4 percentage points, but they were equally supportive of younger and middle-aged candi- dates.19 And when it comes to older voters, we find they are just as critical—if not more—of older candidates as other voters. Respondents 65 and over were 57.0 percentage points less likely to choose older candidates compared to middle-aged candidates, and 55.6 percentage points less likely compared to younger candidates. In sum, contrary to our expectations, we find that neither youth discrimination nor in- group favoritism is likely to play a substantial role in explaining the underrepresentation of young people in elected office. Younger respondents in our experiment strictly preferred younger candidates over others, all age groups preferred younger candidates over older can- didates, and middle-aged and older voters were essentially indifferent between younger and

19While the in-group favoritism results for younger and older voters are robust to alternative age intervals, the results for middle-aged voters are more sensitive. For some age intervals, we find stronger evidence of in- group favoritism, with middle-aged voters as much as 5 percentage points more likely to choose middle-aged candidates over younger candidates.

18 middle-aged candidates.20 The most striking finding is the extent to which respondents of all ages, and elderly voters in particular, disliked elderly-looking candidates. This negative bias based in visual assessment of candidate age is even stronger than the elderly bias found in many conjoint experiments in which candidate ages were listed numerically.21 Our findings thus indicate that voter preferences differ substantially from the observed age demographics of mayors in Japan. Recall from the discussion earlier that the average mayor begins their term at 62 years old. We find no evidence that voters harbor strong negative biases against younger mayoral candidates, yet less than 5% of mayors begin their term before age 45. Younger mayors (aged 25–44) in office are outnumbered 13 to 1 by middle-aged mayors (45–64) and 8 to 1 by older mayors (65 and over).

Age Stereotypes

Why do voters like younger candidates as much as middle-aged candidates and substantially more than older candidates? Why do younger and middle-aged voters prefer candidates closer to their own age, but older voters do not? We turn next to investigating the results of our age stereotype tests concerning policy issues, traits, and electability.

Policy Issues

Figure 3 plots the results for our 11 policy issues. As with our earlier experiments, we average the results across our two models and use Middle-Aged Candidate as the baseline condition.22 We find clear evidence that voters associate different policies with candidates of different ages (H3). Of the 11 policy issues, respondents identified five most closely with younger can- didates. In particular, our survey participants expected younger candidates to be the most likely to stress policies that affect young people, namely, education and childcare. Respon-

20In the Appendix (Figure A1), we find similar results if we re-estimate our analyses using likely voters, that is, those who self-reported that they would turn out for the election. 21For example, Eshima and Smith (2021) find that respondents were over 20 percentage points more likely to say they would vote for a 55-year-old for Japan’s House of Representatives than a 70-year-old. 22Results disaggregated by model and manipulation checks can be found in Tables A4, A5, and A6.

19 Figure 3: Candidate Age and Policy Issues

● Education ●

● Childcare ●

● Climate Change ●

● Anti−Corruption ●

● Foreign Residents ●

● Economy and Employment ●

● Public Works ●

● Budget Deficit ●

● Crime and Safety ●

● Elderly Care ●

● Healthcare ●

● Average (All Issues) ●

−20 −10 0 10 20 Percentage Point Change in Respondent Saying Candidate Is Likely to Emphasize Issue

● Younger Candidate ● Older Candidate

Notes: Middle-Aged Candidate is the baseline category. Bars show 95% confidence intervals. dents said that younger candidates would be 11.0 and 15.5 percentage points more likely to emphasize these two issues than middle-aged candidates, and 21.5 and 26.3 percentage points more likely than older candidates, respectively. Younger candidates were also seen as the most likely to tackle progressive policies such as climate change, anti-corruption measures, and accommodating foreign residents and multiculturalism.

20 When it came to middle-aged candidates, respondents linked four policies concerning the economy and public safety. In a curvilinear relationship, respondents expected middle-aged candidates to focus on the economy and employment, public works, budget deficit, and crime and safety more than either younger or older candidates. These findings for economic policies could reflect voter perceptions that middle-aged candidates strike a balance between having more experience with the economy compared to younger candidates, yet still being active participants in the labor force compared to older candidates. Though it is worth noting that respondents saw younger candidates as only slightly less likely to focus on the economy and employment than middle-aged candidates, and significantly more likely to stress these issues than older candidates. Finally, respondents expected older candidates to focus the most on elderly care and healthcare. Similar to education and childcare for young people, the effect sizes for these welfare policies are large. Respondents believed that older candidates would be 16.8 and 12.9 percentage points more likely to emphasize these issues, respectively, than middle-aged candidates, and 25.0 and 18.5 percentage points more likely than younger candidates. Overall, our respondents indicate that older candidates are seen as devoting the least attention to policy issues of all candidates. Across issues, we can see that respondents were 4.6 percentage points less likely to say older candidates would focus on any given policy issue compared to middle-aged candidates, and 3.9 percentage points less likely compared to younger candidates.

Traits

Figure 4 explores the link between a candidate’s age and respondent evaluations about eight different personal traits.23 In line with our hypothesis (H4), we find that voters expect younger candidates to have less favorable traits than either middle-aged or older candidates on average, although the difference between younger and older candidates is not statistically significant. In general, middle-aged candidates were viewed the most favorably by respon-

23Results disaggregated by model and manipulation checks can be found in Tables A7, A8, and A9.

21 Figure 4: Candidate Age and Traits

Dominant

Reliable

Determined

Considerate

Competent

Consensus−Oriented

Experienced

Long−Term Oriented

Average (All Traits)

−30 −20 −10 0 10 Percentage Point Change in Respondent Saying Candidate Is Likely to Have this Trait

Younger Candidate Older Candidate

Notes: Middle-Aged Candidate is the baseline category. Bars show 95% confidence intervals. dents, with some exceptions. Respondents inferred that middle-age candidates would be more likely to be reliable, determined, competent, and consensus-oriented than candidates from other age groups. Younger candidates were perceived as the most considerate but the least dominant and experienced. And older candidates were judged most experienced but least competent and least long-term oriented. As expected, we find the largest age differences in respondent perceptions concerning candidates’ experience. Respondents were 33.8 percentage points less likely to view younger candidates as experienced compared to older candidates, and 26.7 percentage points less likely relative to middle-aged candidates. There are also sizeable differences with regard to time orientation. While the difference between younger and middle-aged candidates is not

22 Figure 5: Candidate Age and Electability

Electability

−30 −20 −10 0 Percentage Point Change in Respondent Saying Candidate Is Likely to Win Election

Younger Candidate Older Candidate

Notes: Middle-Aged Candidate is the baseline category. Bars show 95% confidence intervals.

statistically significant, both groups are seen as more than 10 percentage points more likely than older candidates to approach policies from a long-term perspective. These results are in line with research that suggests that voters should expect younger politicians to have longer time horizons in office (McClean 2021). Crucially, the average result in Figure 4 suggests that younger candidates are evalu- ated particularly poorly when it comes to personal traits, but this is driven by the outsized effect of experience. If we remove experience, we find that respondents otherwise judge younger candidates as possessing significantly more favorable traits on average than older candidates. For example, respondents viewed younger candidates as more determined, com- petent, consensus-oriented, and long-term oriented than older candidates, though they saw older candidates as more dominant.

Electability

Finally, Figure 5 shows the results for our third age stereotype, electability (H5).24 As with traits, we find mixed results for this hypothesis, as voters judged younger candidates as less electable than middle-aged candidates but more electable than older candidates. When presented with a younger-looking version of a candidate’s photo, respondents were 7.7 percentage points less likely to say the candidate would win an election compared to those

24Results disaggregated by model and manipulation checks can be found in Tables A10, A11, and A12.

23 who viewed the middle-aged version, but 18.9 percentage points more likely compared to those who saw the older version. Taken together, these findings regarding age stereotypes offer interesting insights into our age bias results from Experiment 1. We find support for the strong negative bias against elderly candidates in all three of our Experiment 2 analyses. Respondents viewed older candidates as the least likely to focus on policy issues, least competent, least long-term oriented, and least likely to win an election. Our results indicate that the overall null effect in voter preferences for younger and middle-aged candidates may be a result of different, offsetting mechanisms. Respondents viewed younger candidates as the most likely to focus on many policy issues (including education, childcare, climate change, anti-corruption measures, and foreign residents), but saw middle-aged candidates as the most electable, most likely to focus on the economy and safety, and having the most favorable traits.

Heterogeneous Effects by Respondent Age Group

In Experiment 1, we found that younger and middle-aged respondents, but not older re- spondents, viewed candidates from their age group more favorably than others. In this last section, we test whether participants similarly view candidates from their age group as more likely to emphasize certain issues or traits, or more electable. To do so, we repeat the tests of the previous three sections, but divide the analysis for each test by the age of respondents. Figure 6 shows the results. For space, we report only the averages across all issues and traits here, leaving the breakdown of these categories for the Appendix.25 Overall, we find no clear differences in candidate attribute inferences by respondent age group. Respondents generally agreed about the link between candidate age and issue em- phases, and that middle-aged candidates have more favorable traits and are more electable. We do find a few differences in effect size across age groups, however. Most notably, younger respondents were nearly twice as likely as middle-aged or older respondents to expect younger

25Breakdowns of policies and traits together with manipulation checks can be found in Tables A13–A21.

24 Figure 6: Candidate Age and Attributes by Age of Respondent

Younger Voters Middle−Aged Voters Older Voters (18−44) (45−64) (65 and Over)

Average (All Issues)

Average (All Traits)

Electability

−40 −30 −20 −10 0 10 −40 −30 −20 −10 0 10 −40 −30 −20 −10 0 10 Percentage Point Change in Respondent Saying Candidate Has Given Attribute

Younger Candidate Older Candidate

Notes: Middle-Aged Candidate is the baseline category. Bars show 95% confidence intervals. candidates to emphasize education and childcare (Table A13). Older respondents, surpris- ingly, had the most positive views toward younger candidates on certain issues and traits, including their likelihood of emphasizing economic policies and their competence and relia- bility (Tables A19 and A20). The lack of stronger heterogeneous effects by respondent age group suggests that voters share a general consensus about the relative strengths and weaknesses of candidates of dif- ferent ages. Thus, it is unlikely that the in-group favoritism found in Experiment 1 is driven by differences across voter age groups in expectations about the attributes that younger and older candidates will bring into office. Instead, these age affinity effects may be a result of younger and older voters valuing certain attributes differently. Most clearly, we find signif- icant evidence (Figure 3) that voters expect candidates to focus more on policy issues that are more important for members of the candidate’s own age group. Even though younger voters view younger candidates as less experienced, competent, or electable than middle- aged candidates, younger voters may be relatively happy to support their peers if they value policy attention toward education and childcare more than these personal candidate traits.

25 More research is thus needed to explore how the age of voters affects the relative weight they place on different candidate attributes.

Discussion

Young people are underrepresented in most political institutions. However, our results sug- gest that many voters may want to see this age bias corrected. Across our experimental analyses, we find that voters are equally supportive of younger candidates as middle-aged candidates and that they prefer younger candidates significantly more than older candidates. Our study thus has significant implications for policymakers interested in expanding younger people’s presence in political institutions. We find no evidence that demand-side explanations, such as voter biases, pose significant hurdles to more young people serving in elected office. While more studies are needed to replicate our results and to test the robust- ness of supply-side explanations, the extant evidence suggests that age biases in institutions can best be adjusted by tackling younger people’s lack of political ambition and reforming institutions to make it easier for them to run for office (Joshi 2013; Lawless and Fox 2015; Shames 2017; Stockemer and Sundstr¨om2018). Two caveats to our results are worth mentioning. The first is that our face manipulation experiment simulates the natural changes that come with age, but does not capture efforts to reverse these effects such as undergoing plastic surgery or coloring one’s hair. Many politicians may use these tools to appear younger, and our experiment suggests that voters are likely to reward them for doing so. There is some suggestive, observational evidence that voters in Japan prefer candidates who alter their faces to look younger on their campaign posters (Kohno and Sakurai 2020), and future studies might follow our lead to use face manipulation software in order to test these effects in an experimental setting. The second caveat is that, like many survey experiments, ours was administered online. In other words, our sample is not perfectly representative of the Japanese electorate. It is possible that voters who are willing to participate in online surveys have different views on

26 younger and older mayors than other voters, and these effects may also differ across voter age groups. We do our best to address these concerns by balancing our subject pool to match the latest census. However, future studies might fruitfully test whether voters who do and do not engage with online surveys differ in their attitudes toward the age of politicians. Ultimately, our results add new dimensions to the growing literature on age biases and electoral behavior. Our evidence in experimental analysis of in-group favoritism is similar to the observational results from Webster and Pierce (2019) and Sevi (2021), though we depart in finding that in-group favoritism does not hold for older voters, who tend to be more critical of older candidates than are younger or middle-aged voters. Our findings also contribute to our understanding of the results of earlier, conjoint anal- yses of age discrimination. For instance, our study of a hypothetical mayoral race revealed voter preferences more evenly balanced between younger and middle-aged candidates and much stronger negative biases against older candidates compared to conjoint studies of hypo- thetical candidates for Japan’s House of Representatives (Eshima and Smith 2021; Horiuchi, Smith and Yamamoto 2020). The contrasts in these findings across studies may point to specificity in preference for specific offices (we focus on mayors, who are local executives with greater policy discretion than legislators), or they may reflect our study’s focus on perceived age, rather than chronological age, in a lower information setting. These possibilities present fascinating directions for future research disentangling the ways voters’ age biases might af- fect choice differently across government levels, types of positions, levels of policy influence, and information environments. This study is also the first to focus on age stereotypes as predictors of voters’ judgments regarding a candidate’s policy emphases, traits, and electability. In general, across the voter groups in our study, we find that younger candidates may be viewed as less experienced than older politicians, but they are also seen as more likely to emphasize a wide range of issues in office, including education, childcare, climate change, anti-corruption measures, and policies that benefit foreign residents and promote multiculturalism. By contrast, we find a sharp

27 drop-off in voter support for candidates as they become elderly. Respondents described older candidates as less electable and less competent than others, despite their additional years of experience, and suggested they were least likely to focus on policy issues other than those most relevant to elderly voters. We hope that our new experimental design can inspire additional work on age in other countries and electoral contexts. Our study only scratches the surface in terms of what is possible using age regression and progression software. As these applications grow increas- ingly sophisticated and accessible to scholars, researchers should explore the ways individual ages interact with other identity-based discrimination, such as gender and race stereotypes.

28 References

Adida, Claire L. 2015. “Do African Voters Favor Coethnics? Evidence from a Survey Experiment in Benin.” Journal of Experimental Political Science 2(1):1–11.

Aguilar, Rosario, Saul Cunow and Scott Desposato. 2015. “Choice Sets, Gender, and Can- didate Choice in Brazil.” Electoral Studies 39:230–242.

Arnesen, Sveinung, Dominik Duell and Mikael Poul Johannesson. 2019. “Do Citizens Make Inferences From Political Candidate Characteristics When Aiming for Substantive Repre- sentation?” Electoral Studies 57:46–60.

Barnfield, Matthew. 2020. “Think Twice before Jumping on the Bandwagon: Clarifying Concepts in Research on the Bandwagon Effect.” Political Studies Review 18(4):553–574.

Bidadanure, Juliana. 2015. “Better Procedures for Fairer Outcomes: Youth Quotas in Par- liaments.” Intergenerational Justice Review 2(15):40–46.

Burden, Barry C., Yoshikuni Ono and Masahiro Yamada. 2017. “Reassessing Public Support for a Female President.” The Journal of Politics 79(3):1073–1078.

Busemeyer, Marius R., Achim Goerres and Simon Weschle. 2009. “Attitudes towards Redis- tributive Spending in an Era of Demographic Ageing: The Rival Pressures from Age and Income in 14 OECD Countries.” Journal of European Social Policy 19(3):195–212.

Carnes, Nicholas and Noam Lupu. 2016. “Do Voters Dislike Working-Class Candidates? Voter Biases and the Descriptive Underrepresentation of the Working Class.” American Political Science Review 110(4):832–844.

Chattopadhyay, Raghabendra and Esther Duflo. 2004. “Women as Policy Makers: Evidence from a Randomized Policy Experiment in India.” Econometrica 72(5):1409–1443.

29 Clayton, Amanda, Amanda Lea Robinson, Martha C. Johnson and Ragnhild Muriaas. 2019. “(How) Do Voters Discriminate Against Women Candidates? Experimental and Qualita- tive Evidence From Malawi.” Comparative Political Studies 53(3-4):601–630.

Cleveland, Jeanette N. and Genevieve Hollmann. 1990. “The Effects of the Age-Type of Tasks and Incumbent Age Composition on Job Perceptions.” Journal of Vocational Behavior 36(2):181–194.

Curry, James M. and Matthew R. Haydon. 2018. “Lawmaker Age, Issue Salience, and Senior Representation in Congress.” American Politics Research 46(4):567–595.

Dalton, Russell J. 2008. The Good Citizen: How A Younger Generation Is Reshaping Amer- ican Politics. Washington, D.C.: CQ Press.

Dolan, Kathleen. 2004. Voting for Women: How the Public Evaluates Women Candidates. Boulder, CO: Westview Press.

Duverger, Maurice. 1954. Political Parties, Their Organization and Activity in the Modern State. New York: Wiley.

Eshima, Shusei and Daniel M. Smith. 2021. “Just a Number? Voter Evaluations of

Age in Candidate Choice Experiments.” Working Paper. https://ssrn.com/abstract= 3704473.

Flett, Keira, David Clark-Carter, Sarah Grogan and Rachel Davey. 2013. “How effective are physical appearance interventions in changing smoking perceptions, attitudes and be- haviours? A systematic review.” Tobacco Control 22(2):74–79.

Fox, Justin and Kenneth W. Shotts. 2009. “Delegates or Trustees? A Theory of Political Accountability.” The Journal of Politics 71(4):1225–1237.

30 Franck, Rapha¨eland Ilia Rainer. 2012. “Does the Leader’s Ethnicity Matter? Ethnic Fa- voritism, Education, and Health in Sub-Saharan Africa.” American Political Science Re- view 106(2):294–325.

Goerres, Achim. 2009. The Political Participation of Older People in Europe: The Greying of Our Democracies. New York: Palgrave Macmillan.

Hainmueller, Jens, Daniel J. Hopkins and Teppei Yamamoto. 2014. “Causal Inference in Conjoint Analysis: Understanding Multidimensional Choices via Stated Preference Ex- periments.” Political Analysis 22(1):1–30.

Holbein, John B. and D. Sunshine Hillygus. 2020. Making Young Voters: Converting Civic Attitudes into Civic Action. Cambridge: Cambridge University Press.

Horiuchi, Yusaku. 2005. Institutions, Incentives and Electoral Participation in Japan: Cross- Level and Cross-National Perspectives. New York: Routledge.

Horiuchi, Yusaku, Daniel M. Smith and Teppei Yamamoto. 2020. “Identifying Voter Pref- erences for Politicians’ Personal Attributes: A Conjoint Experiment in Japan.” Political Science Research and Methods 8(1):75–91.

Horiuchi, Yusaku, Tadashi Komatsu and Fumio Nakaya. 2012. “Should Candidates Smile to Win Elections? An Application of Automated Face Recognition Technology.” Political Psychology 33(6):925–933.

Huddy, Leonie and Nayda Terkildsen. 1993. “Gender Stereotypes and the Perception of Male and Female Candidates.” American Journal of Political Science 37(1):119–147.

Jacobson, Gary C. 1983. The Politics of Congressional Elections. Boston: Little, Brown & Co.

Joshi, Devin K. 2013. “The Representation of Younger Age Cohorts in Asian Parliaments: Do Electoral Systems Make a Difference?” Representation 49(1):1–16.

31 Kage, Rieko, Frances M. Rosenbluth and Seiki Tanaka. 2019. “What Explains Low Fe- male Political Representation? Evidence From Survey Experiments in Japan.” Politics & Gender 15(2):285–309.

Kam, Cindy D. and Donald R. Kinder. 2012. “Ethnocentrism as a Short-Term Force in the 2008 American Presidential Election.” American Journal of Political Science 56(2):326– 340.

Kirkland, Patricia A. and Alexander Coppock. 2017. “Candidate Choice Without Party Labels: New Insights From Conjoint Survey Experiments.” Political Behavior 40(3):571– 591.

Kissau, Kathrin, Georg Lutz and Jan Rosset. 2012. “Unequal Representation of Age Groups in Switzerland.” Representation 48(1):63–81.

Kohno, Masaru and Misato Sakurai. 2020. “Senkyo posuta no wakatsukuri de, seijika wa dorehodo toku wo suru no ka? (How Much Do Politicians Benefit by Making Them-

selves Appear Young?).” Ronza https://webronza.asahi.com/politics/articles/ 2020033000002.html.

Kweon, Yesola and ByeongHwa Choi. 2021. “Policy Attitudes Towards the Elderly in an Ag- ing Society: Evidence from a Survey Experiment in Japan.” Political Research Quarterly, Forthcoming.

Lampinen, James, Jack D. Arnal, Jennifer Adams, Kady Courtney and Jason L. Hicks. 2012. “Forensic age progression and the search for missing children.” Psychology, Crime & Law 18(4):405–415.

Lawless, Jennifer L. 2015. “Female Candidates and Legislators.” Annual Review of Political Science 18:349–366.

32 Lawless, Jennifer L. and Richard L. Fox. 2015. Running from Office: Why Young Americans Are Turned Off to Politics. Oxford: Oxford University Press.

Lee, Kristen Schultz. 2016. “Conflicting Views on Elder Care Responsibility in Japan.” Social Science Research 57(133-147).

Lewis, Jonathan and Brian J Masshardt. 2002. “Election Posters in Japan.” Japan Forum 14(3):373–404.

Mansbridge, Jane. 1999. “Should Blacks Represent Blacks and Women Represent Women? A Contingent ‘Yes’.” The Journal of Politics 61(3):628–657.

McClean, Charles T. 2021. “Does the Underrepresentation of Young People in Politi-

cal Institutions Matter? Yes.” Working Paper. https://www.charlesmcclean.com/s/ Does-the-Underrepresentation-of-Young-People-in-Political-Institutions-Matter-Yes. pdf.

McGoldrick, Ann E. and James Arrowsmith. 2001. Discrimination by Age: The Organi- zational Response. In Ageism in Work and Employment, ed. Ian Glover and Mohamed Branine. Burlington, VT: Ashgate pp. 75–96.

Ministry of Internal Affairs and Communications. 2020. Statistical Handbook on Japan.

Muramatsu, Naoko and Hiroko Akiyama. 2011. “Japan: Super-Aging Society Preparing for the Future.” The Gerontologist 51(4):425–432.

Norris, Pippa and Ronald Inglehart. 2001. “Women and Democracy: Cultural Obstacles to Equal Representation.” Journal of Democracy 12(3):126–140.

Ono, Yoshikuni and Barry C. Burden. 2019. “The Contingent Effects of Candidate Sex on Voter Choice.” Political Behavior 41(3):583–607.

Ono, Yoshikuni and Masahiko Asano. 2020. “Candidates’ Facial Attractiveness And Electoral Success: Evidence From Japan’s Upper House Elections.” RIETI Working Paper.

33 Ono, Yoshikuni and Masahiro Yamada. 2020. “Do Voters Prefer Gender Stereotypic Candi- dates? Evidence from a Conjoint Survey Experiment in Japan.” Political Science Research and Methods 8(3):477–492.

Perry, Elissa L., Carol T. Kulik and Anne C. Bourhis. 1996. “Moderating Effects of Personal and Contextual Factors in Age Discrimination.” Journal of Applied Psychology 81(6):628– 647.

Phillips, Anne. 1995. The Politics of Presence. Oxford: Oxford University Press.

Pomante, Michael J. and Scot Schraufnagel. 2015. “Candidate Age and Youth Voter Turnout.” American Politics Research 43(3):479–503.

Posthuma, Richard A. and Michael A. Campion. 2009. “Age Stereotypes in the Workplace: Common Stereotypes, Moderators, and Future Research Directions.” Journal of Manage- ment 35(1):158–188.

Rahn, Wendy M. 1993. “The Role of Partisan Stereotypes in Information Processing About Political Candidates.” American Journal of Political Science 37(2):472–496.

Rittenour, Christine E. and Elizabeth L. Cohen. 2016. “Viewing Our Aged Selves: Age Progression Simulations Increase Young Adults’ Aging Anxiety and Negative Stereotypes of Older Adults.” The International Journal of Aging and Human Development 82(4):271– 289.

Sevi, Semra. 2021. “Do Young Voters Vote for Young Leaders?” Electoral Studies 69:1–8.

Shames, Shauna L. 2017. Out of the Running: Why Millenials Reject Political Careers and Why It Matters. New York: New York University Press.

Shockley, Bethany and Justin J. Gengler. 2020. “Social identity and coethnic voting in the Middle East: Experimental evidence from Qatar.” Electoral Studies 67:1–13.

34 Shugart, Matthew Søberg, Melody Ellis Valdini and Kati Suominen. 2005. “Looking for Locals: Voter Information Demands and Personal Vote-Earning Attributes of Legislators under Proportional Representation.” American Journal of Political Science 49(2):437–449.

Smith, Daniel M. 2018. Dynasties and Democracy: The Inherited Incumbency Advantage in Japan. Stanford, CA: Stanford University Press.

Sniderman, Paul M., Richard A. Brody and Philip E. Tetlock. 1991. Reasoning and Choice: Explorations in Political Psychology. Cambridge: Cambridge University Press.

Stockemer, Daniel and Aksel Sundstr¨om.2018. “Age Representation in Parliaments: Can Institutions Pave the Way for the Young?” European Political Science Review 10(3):467– 490.

Tavits, Margit. 2009. “The Making of Mavericks: Local Loyalties and Party Defection.” Comparative Political Studies 42(6):793–815.

Terkildsen, Nayda. 1993. “When White Voters Evaluate Black Candidates: The Processing Implications of Candidate Skin Color, Prejudice, and Self-Monitoring.” American Journal of Political Science 37(4):1032–1053.

Todorov, Alexander, Anesu N. Mandisodza, Amir Goren and Crystal C. Hall. 2005. “Infer- ences of Competence From Faces Predict Election Outcomes.” Science 308(5728):1623– 1626.

Umeda, Michio. 2020. “The Politics of Aging: Age Difference in Welfare Issue Salience in

Japan 1972–2016.” Political Behavior https://doi.org/10.1007/s11109-020-09627-0.

Wattenberg, Martin P. 2007. Is Voting for Young People? New York: Pearson Longman.

Webster, Steven W. and Andrew W. Pierce. 2019. “Older, Younger, or More Similar? The Use of Age as a Voting Heuristic.” Social Science Quarterly 100(3):635–652.

35 Appendix

Contents

1 Candidate Age and Vote Choice 37 1.1 Likely Voters ...... 37 1.2 Model 1 ...... 37 1.3 Model 2 ...... 38 1.4 Manipulation Check ...... 38

2 Candidate Age and Policy Issues 39 2.1 Model 1 ...... 39 2.2 Model 2 ...... 40 2.3 Manipulation Check ...... 41

3 Candidate Age and Traits 42 3.1 Model 1 ...... 42 3.2 Model 2 ...... 43 3.3 Manipulation Check ...... 44

4 Candidate Age and Electability 45 4.1 Model 1 ...... 45 4.2 Model 2 ...... 45 4.3 Manipulation Check ...... 46

5 Candidate Age and Attributes 47 5.1 Younger Voters ...... 47 5.1.1 Policy Issues ...... 47 5.1.2 Traits ...... 48 5.1.3 Manipulation Check ...... 49 5.2 Middle-Aged Voters ...... 50 5.2.1 Policy Issues ...... 50 5.2.2 Traits ...... 51 5.2.3 Manipulation Check ...... 52 5.3 Older Voters ...... 53 5.3.1 Policy Issues ...... 53 5.3.2 Traits ...... 54 5.3.3 Manipulation Check ...... 55

6 Additional Information on FaceApp 56

36 1 Candidate Age and Vote Choice

1.1 Likely Voters

Figure A1: Candidate Age and Vote Choice (Likely Voters)

(Baseline: Middle−Aged Candidate)

Younger Candidate ●

Older Candidate ●

−60 −40 −20 0 20 Percentage Point Change in Probability of Respondent Voting for Candidate

Notes: Bars show 95% confidence intervals.

1.2 Model 1

Table A1: Candidate Age and Vote Choice (Model 1)

DV: Chosen by Respondent All Younger Voters Middle-Aged Voters Older Voters Respondents (18–44) (45–64) (65 and Over) (1) (2) (3) (4) Younger Candidate 1 0.047∗∗ 0.120∗∗∗ −0.013 0.020 (0.023) (0.039) (0.036) (0.049) Older Candidate 1 −0.551∗∗∗ −0.457∗∗∗ −0.656∗∗∗ −0.555∗∗∗ (0.023) (0.037) (0.037) (0.051) Constant 0.683∗∗∗ 0.640∗∗∗ 0.729∗∗∗ 0.691∗∗∗ (0.017) (0.027) (0.027) (0.036) Observations 1,957 765 754 438 R2 0.299 0.253 0.372 0.279

Notes: Middle-Aged Candidate 1 is the baseline category. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

37 1.3 Model 2

Table A2: Candidate Age and Vote Choice (Model 2)

DV: Chosen by Respondent All Younger Voters Middle-Aged Voters Older Voters Respondents (18–44) (45–64) (65 and Over) (1) (2) (3) (4) Younger Candidate 2 0.00001 0.066∗ −0.036 −0.051 (0.024) (0.039) (0.038) (0.050) Older Candidate 2 −0.529∗∗∗ −0.447∗∗∗ −0.576∗∗∗ −0.585∗∗∗ (0.024) (0.040) (0.037) (0.050) Constant 0.667∗∗∗ 0.610∗∗∗ 0.700∗∗∗ 0.706∗∗∗ (0.017) (0.028) (0.026) (0.036) Observations 1,957 765 754 438 R2 0.250 0.204 0.283 0.289

Notes: Middle-Aged Candidate 2 is the baseline category. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

1.4 Manipulation Check

Table A3: Candidate Age and Vote Choice (Manipulation Check)

DV: Chosen by Respondent All Younger Voters Middle-Aged Voters Older Voters Respondents (18–44) (45–64) (65 and Over) (1) (2) (3) (4) Younger Candidate 0.021 0.078∗∗∗ −0.011 −0.019 (0.017) (0.028) (0.027) (0.036) Older Candidate −0.572∗∗∗ −0.515∗∗∗ −0.626∗∗∗ −0.578∗∗∗ (0.017) (0.028) (0.027) (0.036) Constant 0.687∗∗∗ 0.650∗∗∗ 0.715∗∗∗ 0.703∗∗∗ (0.012) (0.020) (0.019) (0.026) Observations 3,596 1,350 1,400 846 R2 0.305 0.278 0.344 0.291

Notes: Results are limited to respondents who passed a manipulation check asking them to choose which candidate appeared older in cases where candidate ages differed. Middle-Aged Candidate is the baseline category. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

38 2 Candidate Age and Policy Issues

2.1 Model 1

Table A4: Candidate Age and Policy Issues (Model 1)

DV: Policy Issues Climate Anti- Foreign Economy and Education Childcare Change Corruption Residents Employment (1) (2) (3) (4) (5) (6) Younger Candidate 1 0.139∗∗∗ 0.199∗∗∗ 0.063∗∗∗ 0.066∗∗∗ 0.055∗∗∗ −0.022 (0.022) (0.022) (0.021) (0.020) (0.019) (0.022) Older Candidate 1 −0.072∗∗∗ −0.068∗∗∗ −0.081∗∗∗ −0.055∗∗∗ −0.117∗∗∗ −0.165∗∗∗ (0.022) (0.021) (0.020) (0.018) (0.016) (0.021) Constant 0.416∗∗∗ 0.362∗∗∗ 0.318∗∗∗ 0.229∗∗∗ 0.216∗∗∗ 0.443∗∗∗ (0.016) (0.015) (0.015) (0.013) (0.013) (0.016) Observations 3,000 3,000 3,000 3,000 3,000 3,000 R2 0.032 0.054 0.016 0.014 0.033 0.023

Public Budget Crime and Elderly Average Works Deficit Safety Care Healthcare (All Issues) (7) (8) (9) (10) (11) (12) Younger Candidate 1 −0.067∗∗∗ −0.014 −0.036∗ −0.084∗∗∗ −0.074∗∗∗ 0.021 (0.021) (0.021) (0.020) (0.022) (0.022) (0.014) Older Candidate 1 −0.097∗∗∗ −0.088∗∗∗ −0.039∗ 0.173∗∗∗ 0.127∗∗∗ −0.044∗∗∗ (0.021) (0.020) (0.020) (0.022) (0.022) (0.013) Constant 0.380∗∗∗ 0.314∗∗∗ 0.282∗∗∗ 0.470∗∗∗ 0.461∗∗∗ 0.354∗∗∗ (0.016) (0.015) (0.014) (0.016) (0.016) (0.010) Observations 3,000 3,000 3,000 3,000 3,000 3,000 R2 0.007 0.007 0.002 0.046 0.028 0.008

Notes: Middle-Aged Candidate 1 is the baseline category in all models. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

39 2.2 Model 2

Table A5: Candidate Age and Policy Issues (Model 2)

DV: Policy Issues Climate Anti- Foreign Economy and Education Childcare Change Corruption Residents Employment (1) (2) (3) (4) (5) (6) Younger Candidate 2 0.082∗∗∗ 0.112∗∗∗ 0.092∗∗∗ 0.028 0.043∗∗ −0.064∗∗∗ (0.022) (0.022) (0.021) (0.018) (0.019) (0.022) Older Candidate 2 −0.138∗∗∗ −0.149∗∗∗ −0.069∗∗∗ −0.040∗∗ −0.132∗∗∗ −0.139∗∗∗ (0.021) (0.021) (0.019) (0.017) (0.016) (0.021) Constant 0.420∗∗∗ 0.382∗∗∗ 0.274∗∗∗ 0.194∗∗∗ 0.215∗∗∗ 0.429∗∗∗ (0.016) (0.016) (0.014) (0.013) (0.013) (0.016) Observations 3,000 3,000 3,000 3,000 3,000 3,000 R2 0.035 0.050 0.022 0.005 0.037 0.014

Public Budget Crime and Elderly Average Works Deficit Safety Care Healthcare (All Issues) (7) (8) (9) (10) (11) (12) Younger Candidate 2 −0.136∗∗∗ −0.043∗∗ −0.069∗∗∗ −0.079∗∗∗ −0.037∗ −0.006 (0.021) (0.020) (0.019) (0.022) (0.022) (0.014) Older Candidate 2 −0.057∗∗∗ −0.058∗∗∗ −0.032 0.163∗∗∗ 0.131∗∗∗ −0.047∗∗∗ (0.022) (0.020) (0.020) (0.022) (0.022) (0.013) Constant 0.402∗∗∗ 0.302∗∗∗ 0.278∗∗∗ 0.458∗∗∗ 0.404∗∗∗ 0.342∗∗∗ (0.016) (0.015) (0.014) (0.016) (0.016) (0.010) Observations 3,000 3,000 3,000 3,000 3,000 3,000 R2 0.014 0.003 0.004 0.041 0.021 0.005

Notes: Middle-Aged Candidate 2 is the baseline category in all models. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

40 2.3 Manipulation Check

Table A6: Candidate Age and Policy Issues (Manipulation Check)

DV: Policy Issues Climate Anti- Foreign Economy and Education Childcare Change Corruption Residents Employment (1) (2) (3) (4) (5) (6) Younger Candidate 0.115∗∗∗ 0.161∗∗∗ 0.091∗∗∗ 0.052∗∗∗ 0.057∗∗∗ −0.038∗∗ (0.017) (0.016) (0.016) (0.014) (0.014) (0.017) Older Candidate −0.112∗∗∗ −0.113∗∗∗ −0.068∗∗∗ −0.050∗∗∗ −0.122∗∗∗ −0.161∗∗∗ (0.017) (0.016) (0.015) (0.014) (0.012) (0.017) Constant 0.416∗∗∗ 0.371∗∗∗ 0.287∗∗∗ 0.210∗∗∗ 0.210∗∗∗ 0.435∗∗∗ (0.013) (0.012) (0.012) (0.011) (0.010) (0.013) Observations 5,434 5,434 5,434 5,434 5,434 5,434 R2 0.036 0.055 0.021 0.011 0.037 0.019

Public Budget Crime and Elderly Average Works Deficit Safety Care Healthcare (All Issues) (7) (8) (9) (10) (11) (12) Younger Candidate −0.096∗∗∗ −0.024 −0.049∗∗∗ −0.083∗∗∗ −0.055∗∗∗ 0.012 (0.016) (0.015) (0.014) (0.017) (0.016) (0.010) Older Candidate −0.084∗∗∗ −0.079∗∗∗ −0.037∗∗ 0.166∗∗∗ 0.129∗∗∗ −0.048∗∗∗ (0.017) (0.015) (0.015) (0.017) (0.017) (0.010) Constant 0.388∗∗∗ 0.305∗∗∗ 0.278∗∗∗ 0.469∗∗∗ 0.437∗∗∗ 0.346∗∗∗ (0.013) (0.012) (0.011) (0.013) (0.013) (0.008) Observations 5,434 5,434 5,434 5,434 5,434 5,434 R2 0.008 0.005 0.002 0.044 0.024 0.008

Notes: Respondents are limited to only those who estimated the Younger Candidate to be under 45, the Middle-Aged Candidate to be between 45 and 64, and the Older Candidate to be 65 or over. Middle-Aged Candidate is the baseline category in all models. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

41 3 Candidate Age and Traits

3.1 Model 1

Table A7: Candidate Age and Traits (Model 1)

DV: Traits Consensus- Dominant Reliable Determined Considerate Competent Oriented (1) (2) (3) (4) (5) (6) Younger Candidate 1 −0.111∗∗∗ −0.006 −0.010 0.039∗∗ 0.020 −0.008 (0.019) (0.018) (0.021) (0.019) (0.020) (0.019) Older Candidate 1 −0.043∗∗ −0.051∗∗∗ −0.139∗∗∗ 0.055∗∗∗ −0.126∗∗∗ −0.067∗∗∗ (0.020) (0.018) (0.020) (0.020) (0.018) (0.018) Constant 0.290∗∗∗ 0.217∗∗∗ 0.345∗∗∗ 0.226∗∗∗ 0.274∗∗∗ 0.242∗∗∗ (0.015) (0.013) (0.015) (0.013) (0.014) (0.014) Observations 3,000 3,000 3,000 3,000 3,000 3,000 R2 0.012 0.003 0.019 0.003 0.023 0.005

Long-Term Average Experienced Oriented (All Traits) (7) (8) (9) Younger Candidate 1 −0.245∗∗∗ 0.006 −0.039∗∗∗ (0.018) (0.020) (0.012) Older Candidate 1 0.057∗∗∗ −0.143∗∗∗ −0.057∗∗∗ (0.022) (0.018) (0.012) Constant 0.350∗∗∗ 0.270∗∗∗ 0.277∗∗∗ (0.015) (0.014) (0.009) Observations 3,000 3,000 3,000 R2 0.085 0.027 0.008

Notes: Middle-Aged Candidate 1 is the baseline category in all models. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

42 3.2 Model 2

Table A8: Candidate Age and Traits (Model 2)

DV: Traits Consensus- Dominant Reliable Determined Considerate Competent Oriented (1) (2) (3) (4) (5) (6) Younger Candidate 2 −0.125∗∗∗ −0.092∗∗∗ −0.121∗∗∗ 0.028 −0.109∗∗∗ −0.092∗∗∗ (0.016) (0.018) (0.019) (0.020) (0.019) (0.019) Older Candidate 2 0.040∗∗ −0.067∗∗∗ −0.079∗∗∗ −0.015 −0.131∗∗∗ −0.099∗∗∗ (0.019) (0.019) (0.020) (0.020) (0.018) (0.019) Constant 0.224∗∗∗ 0.252∗∗∗ 0.300∗∗∗ 0.274∗∗∗ 0.279∗∗∗ 0.291∗∗∗ (0.013) (0.014) (0.015) (0.014) (0.014) (0.015) Observations 3,000 3,000 3,000 3,000 3,000 3,000 R2 0.032 0.009 0.014 0.002 0.020 0.011

Long-Term Number of Experienced Oriented Traits (7) (8) (9) Younger Candidate 2 −0.288∗∗∗ −0.030 −0.104∗∗∗ (0.017) (0.020) (0.012) Older Candidate 2 0.085∗∗∗ −0.119∗∗∗ −0.048∗∗∗ (0.022) (0.018) (0.012) Constant 0.346∗∗∗ 0.275∗∗∗ 0.280∗∗∗ (0.015) (0.014) (0.009) Observations 3,000 3,000 3,000 R2 0.131 0.014 0.027

Notes: Middle-Aged Candidate 2 is the baseline category in all models. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

43 3.3 Manipulation Check

Table A9: Candidate Age and Traits (Manipulation Check)

DV: Traits Consensus- Dominant Reliable Determined Considerate Competent Oriented (1) (2) (3) (4) (5) (6) Younger Candidate −0.113∗∗∗ −0.049∗∗∗ −0.057∗∗∗ 0.034∗∗ −0.043∗∗∗ −0.044∗∗∗ (0.013) (0.014) (0.015) (0.015) (0.015) (0.014) Older Candidate 0.003 −0.075∗∗∗ −0.113∗∗∗ 0.016 −0.135∗∗∗ −0.087∗∗∗ (0.015) (0.014) (0.015) (0.015) (0.014) (0.014) Constant 0.253∗∗∗ 0.236∗∗∗ 0.316∗∗∗ 0.249∗∗∗ 0.278∗∗∗ 0.261∗∗∗ (0.011) (0.011) (0.012) (0.011) (0.012) (0.011) Observations 5,434 5,434 5,434 5,434 5,434 5,434 R2 0.018 0.006 0.010 0.001 0.018 0.007

Long-Term Average Experienced Oriented (All Traits) (7) (8) (9) Younger Candidate −0.269∗∗∗ −0.005 −0.068∗∗∗ (0.014) (0.015) (0.009) Older Candidate 0.064∗∗∗ −0.135∗∗∗ −0.058∗∗∗ (0.017) (0.014) (0.010) Constant 0.351∗∗∗ 0.267∗∗∗ 0.276∗∗∗ (0.012) (0.011) (0.008) Observations 5,434 5,434 5,434 R2 0.110 0.022 0.013

Notes: Respondents are limited to only those who estimated the Younger Candidate to be under 45, the Middle-Aged Candidate to be between 45 and 64, and the Older Candidate to be 65 or over. Middle-Aged Candidate is the baseline category in all models. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

44 4 Candidate Age and Electability

4.1 Model 1

Table A10: Candidate Age and Electability (Model 1)

DV: Electability (1) Younger Candidate 1 0.030 (0.022) Older Candidate 1 −0.260∗∗∗ (0.019) Constant 0.389∗∗∗ (0.016) Observations 3,000 R2 0.079

Notes: Middle-Aged Candidate 1 is the baseline category in both models. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

4.2 Model 2

Table A11: Candidate Age and Electability (Model 2)

DV: Electability (1) Younger Candidate 2 −0.183∗∗∗ (0.021) Older Candidate 2 −0.273∗∗∗ (0.019) Constant 0.414∗∗∗ (0.016) Observations 3,000 R2 0.066

Notes: Middle-Aged Candidate 1 is the baseline category in both models. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

45 4.3 Manipulation Check

Table A12: Candidate Age and Electability (Manipulation Check)

DV: Electability (1) Younger Candidate −0.067∗∗∗ (0.017) Older Candidate −0.272∗∗∗ (0.015) Constant 0.394∗∗∗ (0.013) Observations 5,434 R2 0.063

Notes: Respondents are limited to only those who estimated the Younger Candidate to be under 45, the Middle-Aged Candidate to be between 45 and 64, and the Older Candidate to be 65 or over. Middle-Aged Candidate is the baseline category in both models. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

46 5 Candidate Age and Attributes

5.1 Younger Voters

5.1.1 Policy Issues

Table A13: Candidate Age and Policy Issues (Younger Voters)

DV: Policy Issues Climate Anti- Foreign Economy and Education Childcare Change Corruption Residents Employment (1) (2) (3) (4) (5) (6) Younger Candidate 0.148∗∗∗ 0.201∗∗∗ 0.082∗∗∗ 0.037∗ 0.032 −0.048∗∗ (0.023) (0.023) (0.021) (0.020) (0.021) (0.024) Older Candidate −0.100∗∗∗ −0.096∗∗∗ −0.037∗ −0.051∗∗∗ −0.117∗∗∗ −0.149∗∗∗ (0.022) (0.022) (0.020) (0.018) (0.018) (0.023) Constant 0.412∗∗∗ 0.356∗∗∗ 0.270∗∗∗ 0.206∗∗∗ 0.231∗∗∗ 0.474∗∗∗ (0.017) (0.017) (0.015) (0.014) (0.015) (0.018) Observations 2,774 2,774 2,774 2,774 2,774 2,774 R2 0.043 0.066 0.012 0.008 0.026 0.016

Public Budget Crime and Elderly Average Works Deficit Safety Care Healthcare (All Issues) (7) (8) (9) (10) (11) (12) Younger Candidate −0.134∗∗∗ −0.037∗ −0.059∗∗∗ −0.087∗∗∗ −0.071∗∗∗ 0.006 (0.022) (0.022) (0.020) (0.024) (0.023) (0.014) Older Candidate −0.043∗ −0.042∗ −0.025 0.157∗∗∗ 0.116∗∗∗ −0.035∗∗∗ (0.023) (0.022) (0.020) (0.023) (0.023) (0.013) Constant 0.406∗∗∗ 0.328∗∗∗ 0.279∗∗∗ 0.476∗∗∗ 0.464∗∗∗ 0.355∗∗∗ (0.017) (0.016) (0.016) (0.018) (0.018) (0.011) Observations 2,774 2,774 2,774 2,774 2,774 2,774 R2 0.014 0.002 0.003 0.042 0.024 0.004

Notes: Sample is limited to younger respondents who are under 45 years old. Middle-Aged Candidate is the baseline category in all models. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

47 5.1.2 Traits

Table A14: Candidate Age and Traits (Younger Voters)

DV: Traits Consensus- Dominant Reliable Determined Considerate Competent Oriented (1) (2) (3) (4) (5) (6) Younger Candidate −0.140∗∗∗ −0.049∗∗ −0.100∗∗∗ 0.050∗∗ −0.077∗∗∗ −0.064∗∗∗ (0.020) (0.020) (0.021) (0.021) (0.021) (0.020) Older Candidate −0.024 −0.054∗∗∗ −0.118∗∗∗ 0.014 −0.130∗∗∗ −0.056∗∗∗ (0.021) (0.020) (0.022) (0.021) (0.020) (0.020) Constant 0.284∗∗∗ 0.247∗∗∗ 0.359∗∗∗ 0.265∗∗∗ 0.310∗∗∗ 0.271∗∗∗ (0.015) (0.015) (0.017) (0.015) (0.016) (0.016) Observations 2,774 2,774 2,774 2,774 2,774 2,774 R2 0.021 0.003 0.013 0.002 0.015 0.004

Long-Term Average Experienced Oriented (All Traits) (7) (8) (9) Younger Candidate −0.327∗∗∗ −0.013 −0.090∗∗∗ (0.019) (0.021) (0.013) Older Candidate 0.069∗∗∗ −0.113∗∗∗ −0.051∗∗∗ (0.023) (0.020) (0.013) Constant 0.395∗∗∗ 0.293∗∗∗ 0.303∗∗∗ (0.018) (0.016) (0.011) Observations 2,774 2,774 2,774 R2 0.142 0.014 0.019

Notes: Sample is limited to younger respondents who are under 45 years old. Middle-Aged Candidate is the baseline category in all models. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

48 5.1.3 Manipulation Check

Table A15: Candidate Age and Attributes (Younger Voters, Manipulation Check)

Dependent Variable Average (All Issues) Average (All Traits) Electability (1) (2) (3) Younger Candidate 0.015 −0.087∗∗∗ −0.078∗∗∗ (0.014) (0.014) (0.025) Older Candidate −0.032∗∗ −0.057∗∗∗ −0.245∗∗∗ (0.014) (0.015) (0.024) Constant 0.347∗∗∗ 0.300∗∗∗ 0.421∗∗∗ (0.012) (0.012) (0.019) Observations 2,456 2,456 2,456 R2 0.005 0.019 0.046

Notes: Sample is limited to younger respondents who are under 45 years old and passed a manipulation check where they estimated the Younger Candidate to be under 45, the Middle-Aged Candidate to be between 45 and 64, and the Older Candidate to be 65 or over. Middle-Aged Candidate is the baseline category. Standard errors are clustered by respondent. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

49 5.2 Middle-Aged Voters

5.2.1 Policy Issues

Table A16: Candidate Age and Policy Issues (Middle-Aged Voters)

DV: Policy Issues Climate Anti- Foreign Economy and Education Childcare Change Corruption Residents Employment (1) (2) (3) (4) (5) (6) Younger Candidate 0.067∗∗ 0.107∗∗∗ 0.056∗∗ 0.060∗∗ 0.038∗ −0.075∗∗∗ (0.028) (0.028) (0.027) (0.024) (0.022) (0.027) Older Candidate −0.140∗∗∗ −0.133∗∗∗ −0.131∗∗∗ −0.047∗∗ −0.121∗∗∗ −0.187∗∗∗ (0.027) (0.025) (0.025) (0.022) (0.019) (0.026) Constant 0.412∗∗∗ 0.370∗∗∗ 0.313∗∗∗ 0.199∗∗∗ 0.188∗∗∗ 0.425∗∗∗ (0.021) (0.020) (0.019) (0.017) (0.016) (0.021) Observations 1,956 1,956 1,956 1,956 1,956 1,956 R2 0.032 0.042 0.030 0.012 0.034 0.026

Public Budget Crime and Elderly Average Works Deficit Safety Care Healthcare (All Issues) (7) (8) (9) (10) (11) (12) Younger Candidate −0.115∗∗∗ −0.032 −0.054∗∗ −0.087∗∗∗ −0.045 −0.007 (0.026) (0.025) (0.024) (0.027) (0.027) (0.018) Older Candidate −0.118∗∗∗ −0.109∗∗∗ −0.062∗∗ 0.145∗∗∗ 0.121∗∗∗ −0.071∗∗∗ (0.027) (0.024) (0.024) (0.028) (0.028) (0.016) Constant 0.389∗∗∗ 0.289∗∗∗ 0.264∗∗∗ 0.449∗∗∗ 0.395∗∗∗ 0.336∗∗∗ (0.020) (0.019) (0.019) (0.021) (0.021) (0.014) Observations 1,956 1,956 1,956 1,956 1,956 1,956 R2 0.014 0.011 0.004 0.037 0.020 0.011

Notes: Sample is limited to middle-aged respondents between 45 and 64 years old. Middle-Aged Candidate is the baseline category in all models. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

50 5.2.2 Traits

Table A17: Candidate Age and Traits (Middle-Aged Voters)

DV: Traits Consensus- Dominant Reliable Determined Considerate Competent Oriented (1) (2) (3) (4) (5) (6) Younger Candidate −0.115∗∗∗ −0.055∗∗ −0.037 0.018 −0.027 −0.048∗∗ (0.021) (0.022) (0.025) (0.024) (0.024) (0.024) Older Candidate −0.004 −0.076∗∗∗ −0.104∗∗∗ 0.009 −0.144∗∗∗ −0.105∗∗∗ (0.024) (0.022) (0.024) (0.023) (0.021) (0.022) Constant 0.239∗∗∗ 0.207∗∗∗ 0.276∗∗∗ 0.223∗∗∗ 0.246∗∗∗ 0.245∗∗∗ (0.017) (0.017) (0.019) (0.017) (0.018) (0.018) Observations 1,956 1,956 1,956 1,956 1,956 1,956 R2 0.018 0.007 0.010 0.0003 0.025 0.012

Long-Term Average Experienced Oriented (All Traits) (7) (8) (9) Younger Candidate −0.218∗∗∗ −0.027 −0.064∗∗∗ (0.022) (0.024) (0.015) Older Candidate 0.040 −0.155∗∗∗ −0.067∗∗∗ (0.027) (0.021) (0.014) Constant 0.305∗∗∗ 0.256∗∗∗ 0.250∗∗∗ (0.019) (0.018) (0.012) Observations 1,956 1,956 1,956 R2 0.072 0.029 0.015

Notes: Sample is limited to middle-aged respondents between 45 and 64 years old. Middle-Aged Candidate is the baseline category in all models. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

51 5.2.3 Manipulation Check

Table A18: Candidate Age and Attributes (Middle-Aged Voters, Manipulation Check)

Dependent Variable Average (All Issues) Average (All Traits) Electability (1) (2) (3) Younger Candidate −0.007 −0.059∗∗∗ −0.037 (0.019) (0.016) (0.028) Older Candidate −0.078∗∗∗ −0.067∗∗∗ −0.257∗∗∗ (0.018) (0.016) (0.024) Constant 0.339∗∗∗ 0.246∗∗∗ 0.340∗∗∗ (0.015) (0.013) (0.021) Observations 1,806 1,806 1,806 R2 0.013 0.013 0.068

Notes: Sample is limited to middle-aged respondents who are between 45 and 64 years old and passed a manipulation check where they estimated the Younger Candidate to be under 45, the Middle-Aged Candidate to be between 45 and 64, and the Older Candidate to be 65 or over. Middle-Aged Candidate is the baseline category. Standard errors are clustered by respondent. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

52 5.3 Older Voters

5.3.1 Policy Issues

Table A19: Candidate Age and Policy Issues (Older Voters)

DV: Policy Issues Climate Anti- Foreign Economy and Education Childcare Change Corruption Residents Employment (1) (2) (3) (4) (5) (6) Younger Candidate 0.098∗∗∗ 0.133∗∗∗ 0.102∗∗∗ 0.050 0.103∗∗∗ 0.016 (0.033) (0.034) (0.033) (0.031) (0.031) (0.033) Older Candidate −0.060∗ −0.093∗∗∗ −0.069∗∗ −0.039 −0.149∗∗∗ −0.111∗∗∗ (0.035) (0.033) (0.032) (0.029) (0.026) (0.032) Constant 0.438∗∗∗ 0.407∗∗∗ 0.326∗∗∗ 0.244∗∗∗ 0.225∗∗∗ 0.375∗∗∗ (0.025) (0.025) (0.024) (0.023) (0.022) (0.024) Observations 1,270 1,270 1,270 1,270 1,270 1,270 R2 0.017 0.035 0.022 0.007 0.064 0.014

Public Budget Crime and Elderly Average Works Deficit Safety Care Healthcare (All Issues) (7) (8) (9) (10) (11) (12) Younger Candidate −0.012 −0.006 −0.034 −0.062∗ −0.038 0.032 (0.033) (0.031) (0.030) (0.034) (0.034) (0.023) Older Candidate −0.095∗∗∗ −0.095∗∗∗ −0.017 0.226∗∗∗ 0.166∗∗∗ −0.030 (0.034) (0.030) (0.033) (0.035) (0.035) (0.021) Constant 0.363∗∗∗ 0.295∗∗∗ 0.307∗∗∗ 0.461∗∗∗ 0.424∗∗∗ 0.351∗∗∗ (0.025) (0.024) (0.024) (0.026) (0.026) (0.018) Observations 1,270 1,270 1,270 1,270 1,270 1,270 R2 0.008 0.010 0.001 0.061 0.031 0.007

Notes: Sample is limited to older respondents who are 65 years old or older. Middle-Aged Candidate is the baseline category in all models. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

53 5.3.2 Traits

Table A20: Candidate Age and Traits (Older Voters)

DV: Traits Consensus- Dominant Reliable Determined Considerate Competent Oriented (1) (2) (3) (4) (5) (6) Younger Candidate −0.077∗∗∗ −0.038 −0.039 0.024 −0.004 −0.020 (0.027) (0.030) (0.031) (0.030) (0.030) (0.029) Older Candidate 0.049 −0.043 −0.104∗∗∗ 0.052 −0.107∗∗∗ −0.113∗∗∗ (0.031) (0.031) (0.031) (0.032) (0.029) (0.029) Constant 0.230∗∗∗ 0.251∗∗∗ 0.316∗∗∗ 0.260∗∗∗ 0.253∗∗∗ 0.286∗∗∗ (0.021) (0.023) (0.023) (0.022) (0.022) (0.023) Observations 1,270 1,270 1,270 1,270 1,270 1,270 R2 0.016 0.002 0.009 0.002 0.014 0.013

Long-Term Average Experienced Oriented (All Traits) (7) (8) (9) Younger Candidate −0.215∗∗∗ 0.013 −0.044∗∗ (0.027) (0.030) (0.019) Older Candidate 0.116∗∗∗ −0.138∗∗∗ −0.036∗ (0.034) (0.028) (0.019) Constant 0.316∗∗∗ 0.255∗∗∗ 0.271∗∗∗ (0.024) (0.022) (0.015) Observations 1,270 1,270 1,270 R2 0.093 0.027 0.005

Notes: Sample is limited to older respondents who are 65 years old or older. Middle-Aged Candidate is the baseline category in all models. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

54 5.3.3 Manipulation Check

Table A21: Candidate Age and Attributes (Older Voters, Manipulation Check)

Dependent Variable Average (All Issues) Average (All Traits) Electability (1) (2) (3) Younger Candidate 0.034 −0.045∗∗ −0.088∗∗ (0.024) (0.019) (0.035) Older Candidate −0.037 −0.048∗∗ −0.355∗∗∗ (0.023) (0.021) (0.031) Constant 0.354∗∗∗ 0.274∗∗∗ 0.423∗∗∗ (0.019) (0.016) (0.027) Observations 1,172 1,172 1,172 R2 0.009 0.007 0.109

Notes: Sample is limited to older respondents who are 65 years old or older and passed a manipulation check where they estimated the Younger Candidate to be under 45, the Middle-Aged Candidate to be between 45 and 64, and the Older Candidate to be 65 or over. Middle-Aged Candidate is the baseline category. Standard errors are clustered by respondent. ∗p<.1; ∗∗p<.05; ∗∗∗p<.01.

55 6 Additional Information on FaceApp

FaceApp is a mobile application created by Wireless Lab that allows users to realistically manipulate characteristics of faces in photos, including age. The app has both free and paid tiers.26 Initially released in January 2017, the app became especially popular in the summer of 2019 because of its photo-realism and its use by many celebrities. On social media, the app went viral on Twitter and Instagram under the #AgeChallenge, which challenged people to upload images of themselves with the app’s old-age filter applied. By July 2019, more than 150 million people had downloaded the app. However, the app also came under controversy at the time because of privacy concerns about how the Russian company was using an individual’s photos.27 While FaceApp uses a proprietary algorithm that Wireless Lab does not share publicly, the company has said that it ages or de-ages photos using artificial intelligence and neural networks. More specifically, this process involves manipulating faces along several dimen- sions, including changes to (i) wrinkles, especially on the forehead, above the nose, and in the smile lines between the nose and mouth; (ii) skin elasticity, as skin becomes looser with age, especially underneath the eyelids and around the neck; (iii) color contrast, as faces with high color contrast between the eyes, lip, and mouth tend to appear younger than faces with low contrast; (iv) skin pigmentation, as hormones and sun exposure darken the skin over time; and finally (v) hair color, as our hair follicles tend to grow grey, silver, or white with age.28

26FaceApp, https://www.faceapp.com, accessed August 1, 2020. 27See for example John Koetsier, “Viral App FaceApp Now Owns Access to More Than 150 Million People’s Faces and Names,” Forbes, July 17, 2019,https://www.forbes.com/sites/johnkoetsier/2019/07/17/viral-app- faceapp-now-owns-access-to-more-than-150-million-peoples-faces-and-names/10b33ede62f1. 28For more information on FaceApp, see Vishal Thakur, “How Does FaceApp Work?” Sci- ence ABC, January 10, 2020, https://www.scienceabc.com/innovation/how-does-faceapp-work.html; Alexander Kacala, “FaceApp Challenge: Everything You Need to Know About the App People Are Using to Age Themselves,” Newsweek, January 14, 2019, https://www.newsweek.com/faceapp- challenge-everything-you-need-know-about-app-people-are-using-age-themselves-1449158; and Natasha Lo- mas, “FaceApp Uses Neural Networks for Photorealistic Selfie Tweaks,” TechCrunch, February 8, 2017, https://techcrunch.com/2017/02/08/faceapp-uses-neural-networks-for-photorealistic-selfie-tweaks/.

56