Apart from their careers in politics, Mitch McConnell, Joe Biden, and Nancy Pelosi have a least one other thing in common: they are all over age 75. Biden takes the title as the oldest president in history, and Pelosi is the oldest to serve as Speaker of the House. As leaders, they govern alongside a congress that is the oldest in modern memory.Footnote 1 Baby boomers compose about a quarter of the electorate, but members of this generation hold over half of the seats in the House of Representatives and more than two thirds of the seats in the Senate. While about 37% of the American electorate is aged 40 or younger, only 7% of seats in 117th Congress are held by legislators in the same age group (Blazina and Silver 2021; Cilluffo and Fry 2019). The same patterns travel to state legislatures, where the average lawmaker is about a decade older than the average age of American adults (Kurtz, 2015). Relative the age profile of the American electorate, older voices are overrepresented within the federal government.

Is this consistent with the preferences of the electorate? It may be that voters prefer older politicians, perceiving younger candidates as less experienced and less qualified to serve. Alternately, people may dislike older politicians, perceiving them as less mentally sharp and less effective at representing their interests. Rather than reflecting the desires of the electorate, the over-representation of older politicians in government may be imposed by structural factors such as incumbency and candidate recruitment that favors established senior politicians. We explore whether people prefer older or younger politicians to represent their interests in government, as well as differences in the preferences of older and younger people. Even if younger people say they would like to see more candidates like them (Astor, 2020; Taylor, 2018), they remain less likely to turn out to vote than older cohorts (Rouse & Ross, 2018). If older voters prefer older politicians who resemble them, the current age profile of elected officials may well be consistent with the demands of the voting electorate.

Our research speaks to public demand for descriptive representation by age. If people prefer younger politicians because they believe that they can represent their interests in a way that older politicians cannot, then the current age profile of politicians in elected office will fail to represent the public’s preferences for what kinds of people should represent them. Our research also speaks to the importance of age as a political category. Like gender, race, ethnicity, and religion, age is a dimension that people use to categorize both others and themselves (Fiske, 1998). Yet evidence seems mixed as to whether age is necessarily a salient category in politics. When asked about their candidate preferences, Americans say they prefer middle-aged candidates. In a May 2019 survey by the Pew Research Center, about half of Americans said that the best age for a president is in their 50s. However, it is less clear that voters act on these preferences in practice. Consider the case of Joe Biden, who emerged as the Democratic nominee for president in 2020 from a field of candidates that included mostly younger competitors.

To explore whether people evaluate younger and older politicians differently from middle-aged politicians, we first use an experiment to consider the effects of candidate age in evaluations of state legislative candidates. We find that people do not seem to use age as a main criterion in evaluating candidates for office. In our experiment, people are equally supportive of a 23-year-old candidate, a 50-year-old candidate, and a 77-year-old candidate. However, we find that younger candidates for the state legislature are more likely to be seen as inexperienced and unqualified for office. We also find that a politician’s age can serve as an ideological cue, where younger candidates are less likely to be seen as conservative. These stereotypes can affect whether younger candidates are perceived as viable candidates for office.

We then turn to survey evidence to explore whether people evaluate the job performance of their representatives in Congress differently as a function of their legislator’s age. In our observational data, we find that a politician’s increasing age has a small but statistically significant negative effect on approval of lawmakers in Congress. Older legislators enjoy lower levels of approval than younger lawmakers, even after controlling for differences in the productivity and ideology of representatives. Moreover, the penalties for advanced age are doled out especially from younger constituents as compared to older ones. This is consistent with arguments that young people are frustrated by their lack of descriptive representation in politics (Astor, 2020; Taylor, 2018). While we fail to find that younger people have a preference for younger candidates, they are unhappy with the older politicians who represent them in Congress.

Why are young people underrepresented in political office?

In thinking about the reasons why younger people are under-represented in politics, this is in part by design. The Constitution sets minimum age requirements for the presidency, the House of Representatives, and the Senate. Most states have enacted similar rules, requiring candidates for the governorship and state legislature to be several years past the voting age to be eligible to run. Because serving in the state legislature can be a first stepping stone in a political career, these rules can reinforce the underrepresentation of younger people in other corners of politics as well.

Once eligible to run, young people may find it harder to become successful candidates. Younger people are less likely to have accumulated the kind of experiences that can help them stand out from other candidates, and the costs of running a campaign can seem daunting to young people considering a career in politics (Shames, 2017). Political party leaders can hold biases about which people are viable candidates to win elections for the party (Doherty et al., 2019), though some work suggests that age does not predict being recruited to run for office (Fox & Lawless, 2010). The supply of younger candidates for office is also limited by young people themselves. When young people are asked about their interest in pursuing political careers, they tend to be reluctant, turned off by the acrimony of politics and the challenges of modern campaigns (Lawless & Fox, 2015; Shames, 2017).

Beyond the structural factors that limit younger people’s emergence as political candidates, are they also held back by voters who are unwilling to support their candidacies when they run? We have reason to believe that the age biases held by voters could impose barriers to young people’s political candidacies. Restricting young people’s access to legislative office may be an accepted aspect of the design of government. The framers of the Constitution placed restrictions on young people’s political eligibility with little discussion and limited dissent (Nwanevu, 2014). When given an opportunity to make political careers more accessible for young people, nearly two-thirds of Colorado voters rejected a 2018 ballot initiative that would have lowered the age of serving in the state legislature from 25 to 21.

In surveys, people reveal their doubts about young people’s suitability to serve in elected office. In February 2020, 40% of Americans admitted that they would have some reservations about supporting a presidential candidate under the age of forty.Footnote 2 In studies of candidate appearance, older candidates are seen as more mature and more competent, and perceptions of being older predict both perceptions of competence and winning elections (Herrick et al., 2012; Olivola and Todorov 2010). Political experience is often valued in candidates (Kirkland and Coppock 2018; Teele, Kalla, and Rosenbluth 2018), but is less likely to be possessed by younger candidates.

Yet when thinking about how age is stereotyped in politics, being older can be both a positive and a negative. Older people can be perceived as distinguished and wise, but they can also be seen as out of touch and slow to adapt (Hummert, 2011). Advanced age can be associated with lower levels of competence, as reflected in stereotypes about rambling speech, closed-mindedness, forgetfulness, and declining mental sharpness (Hummert et al., 1994). These doubts have led some to recommend tests of mental acuity and cognitive decline for elected officials over the age of 75 (Goldberg, 2017). In contrast, youth is attached to positive stereotypes of attractiveness and liveliness (Kite et al., 2005). Outside of politics, young is typically seen as preferable to old, and negative stereotypes about advanced age outnumber positive ones (Fiske et al., 2002; Kite et al., 2005). Many admit to explicit biases against older people, and most hold implicit biases that favor young people over older people (Chopik & Giasson, 2017; Nosek, Banaji, and Greenwald 2002). These implicit biases occur across age groups, and may be greater in magnitude than the implicit biases that white people hold toward African Americans (Chopik & Giasson, 2017; Nosek, Banaji, and Greenwald 2002).

These biases about older people can travel into politics. Studies in psychology on the relative effects of sexism, racism, and ageism in candidate evaluation suggest that age biases may matter more than gender or racial biases. In an experiment conducted with sample of college students, Sigelman & Sigelman (1982) find a clear preference for a younger 31-year-old candidate running against a middle-aged opponent, as well as a preference for the middle-aged candidate when he competes against a 71-year-old opponent. However, in the same study, participants generally do not seem to prefer male candidates to female candidates or white candidates to black candidates. In a later experiment with a similar research design and a more diverse sample of participants, Piliavin (1987) also finds effects associated with candidate age, but not race or gender. In this study, people are less supportive of the older candidate compared to either a middle-aged or younger candidate, but do not show a preference for the young candidate over the middle-aged candidate. This may reflect a greater social acceptability of acting on age biases over those associated with race, ethnicity, or gender.

Yet even as psychologists find that ageism is more consequential for candidate selection than a candidate’s race or gender, the effects of age in politics have received far less attention than the effects of race and gender. To the extent to which age arises, it is often in the context of presidential campaigns. When candidate age comes up in campaign discourse, its relative salience varies. When Ronald Reagan ran for president in 1980, he seemed able to mostly deflect critiques about whether he was too old to serve as president at age 69 (Kenski, Hardy, and Jamieson 2010). Bob Dole faced media scrutiny about his age (73) during his bid for the presidency in 1996, yet most voters said that his age would not be a barrier to him being an effective president (Abrams and Brody 1998). In 2008, many said that they believed that 72-year-old John McCain was too old to serve effectively as president. These views proved consequential, as those who felt this were less likely to support him, even after controlling for partisan differences (Kenski, Hardy, and Jamieson 2010). However, when 69-year-old Hillary Clinton battled 70-year-old Donald Trump in the 2016 presidential race, the candidates’ ages seemed to draw minimal attention during the campaign. Likewise, the candidates’ ages were not a central campaign focus in 2020 when 74-year-old Donald Trump battled 77-year-old Joe Biden. Perhaps candidate age is only salient when in cases where meaningful age gaps emerge between candidates and their opponents.

To the extent to which political scientists have explored the consequences of candidate age for voter evaluations, it is most commonly through conjoint experiments that vary a number of candidate demographic attributes to see how these variations predict vote choice. In the simplest case where people are given only information about candidates’ age and gender, Klofstad, Anderson, and Nowicki (2015) find that candidates in their 40s and 50s perform better in mock elections than those in their 30s or 60s. Candidates in their 70s are rarely favored against candidates of any younger age. When asked to consider candidates’ age alongside a wider range of demographic differences, people do not seem as likely to differentiate between a range of middle-aged candidates. Hainmueller, Hopkins, and Yamamoto (2014) find that experimental participants similarly rate candidates between the ages of 35 and 60, though they do prefer younger candidates to a 75-year-old politician. Interestingly, they confirm the same result as Sigelman and Sigelman (1982) and Piliavin (1987), where candidate age is more predictive of candidate choice than the gender or race of the candidate.

Other conjoint experiments vary candidate age alongside more substantive attributes like partisanship and political experience, and find more modest effects associated with age. Among these studies, some confirm penalties for candidates aged 70 and older but find modest differences across other ages (Bansak et al., 2021; Ono & Burden, 2019). Less commonly, some studies suggest that 45-year-old candidates are preferred to both younger and older candidates (Kirkland and Coppock 2018; Teele, Kalla, and Rosenbluth 2018).

In the context of conjoint experiments, people are asked to compare two candidates of different ages, where one candidate’s age is evaluated against the benchmark of the other candidate’s age. However, people do not always learn about candidates in explicitly comparative ways. From political psychology, we know that people process political information differently when it focuses on a single candidate, such as in a campaign ad, than in side-by-side comparisons, like a candidate debate (Rahn, Aldrich, and Borgida 1994). Conjoint experiments can speak to how people compare alternatives on a ballot, but questions remain about how a candidate’s age informs the impressions people form of specific politicians in other settings. In our analysis, we consider the effects of candidate age outside the specifically comparative assessment captured in conjoint analysis. We expand beyond vote choice to consider the effects of candidate age on several kinds of evaluations, including favorability toward the candidate, perceptions of how well the candidate can represent the needs of the district, and the stereotypes people associate with younger and older candidates.

We also extend prior studies by considering heterogeneity in the effects of candidate age by the age of the respondent, something rarely considered in past studies of the effects of age. McDonald and Deckman (forthcoming) speak to this in part in their analysis of a sample of 18–22 year olds, but they ultimately find little difference in how a sample of Generation Z respondents evaluate a 22-year-old candidate compared to a 68-year-old one. Webster and Pierce (2019) consider whether voters rely on a heuristic of selecting a candidate who is closest to them in age. But they find only partial support for this expectation. In 2010, partisans are less likely to support their party’s nominee in the election as the age gap between the candidate and respondent increases. However, this pattern does not repeat in their analysis of vote choice in 2012, and in some cases, the reverse pattern is found.Footnote 3

We anticipate that people will tend to give lower ratings of both younger candidates and older candidates. However, we expect these biases to be conditional on the age of the respondent. We expect young people will give more enthusiastic ratings of younger politicians compared to older politicians. In surveys, young people say that they are disappointed by their representation by older politicians. Only 16% of a sample of 18 to 29-year-olds felt that elected officials who are part of the baby boomer generation cared about people like them.Footnote 4 Most young people say they would like to see younger politicians. In a 2018 sample of 15 to 34-year-olds, four out of five agreed, “New leaders from my generation would do a better job running the country.”Footnote 5 To the extent to which young people value descriptive representation by age, we should see the most negative ratings of older politicians among the youngest cohorts.

We expect that older people will give more negative ratings of younger politicians and warmer evaluations of older politicians. Just as younger people may have an interest in electing people who descriptively share the same generational experiences as themselves, the same may well be true for older generations. Older people are often thought to hold distinct group interests in politics, connected to shared concerns like Social Security and protecting health care benefits (Campbell and Binstock 2010; MacManus 1995). These group interests have been shown to mobilize older voters to action (Campbell, 2003). As such, older people may prefer older representatives, believing that politicians closer in age may better appreciate the unique needs of senior citizens and older voters generally.

Experimental study of politician age and candidate preferences

To consider people’s preferences for candidates who are younger or older, we first rely on an experiment placed on a module of the 2020 Cooperative Election Study.Footnote 6 Our sample includes 1000 respondents who completed the experiment online in the course of a pre-election wave of the survey.Footnote 7 Experimental participants were given a short vignette about a candidate running for the state legislature, selected as a venue where many start their political career. The text of the treatment is provided in the supplemental appendix. All participants were given the same background details about the candidate’s profession, alma mater, issue priorities, and reasons for running for office. We include these substantive details about the candidate because we wanted to assess the consequence of candidate age in a setting where age is only one of several details about a candidate. This not only better reflects the way people assess candidates in real life, but it also enhances the realism of the evaluative task for our participants. By mirroring the kind of biography a person might encounter in the news, we hope to promote internal validity, as people take on the task in same ways they might evaluate the candidate in real life.

In the first treatment condition, the candidate was described as 23 years old. In the second treatment condition, the candidate was described as 77 years old. In the control condition, the candidate was described as aged fifty. We selected these ages in part to provide a sharp signal of the ages of the candidate, to ensure that the treatment was received by the participants. To further ensure that the age treatment was received by participants, the last sentence primed age again by mentioning that if elected, the candidate would be either the youngest or oldest lawmaker in the state legislature.Footnote 8 In many past studies of candidate age (e.g. Hainmueller, Hopkins, and Yamamoto 2014; Kirkland and Coppock 2018), the youngest candidate is set at age 35, the minimum age to serve as president. We extend past studies by considering support for a candidate who is much younger, landing between the modal age threshold to serve in the lower house of state legislatures (21) and the minimum age to serve in the U.S. House of Representatives (25).Footnote 9 As a result, we are able to assess whether people are reluctant to support significantly younger candidates just starting a career in politics.

After reading the vignette, we ask respondents to evaluate how likely it is that they would vote for this candidate on a scale from 0 (very unlikely) to 100 (very likely). In Fig. 1, we summarize average levels of candidate support by condition.Footnote 10 We find little evidence that people are biased against younger or older candidates. In the control condition, the mean likelihood of supporting the candidate is 55.0, which is statistically indistinguishable from the level of support voiced by those who read about the 23-year-old candidate in a two-tailed t-test (mean = 54.6, t = 0.22, p < 0.83).Footnote 11 Those who read about the 77-year-old candidate were somewhat less likely to express support for the politician compared to those in the control condition, but the difference is not significant at conventional levels (mean = 52.0, t = 1.75, p < 0.08).Footnote 12

Fig. 1
figure 1

Support for Older versus Younger Candidate, Experimental Results

Full size image

In contrast to conjoint studies that find that people are reluctant to back candidates over age 70 when pitted against younger opponents, we fail to find evidence that people are particularly unwilling to support a 77-year-old candidate when given a vignette that describes his background and platform. Likewise, we fail to find evidence that people are less willing to support a 23-year-old candidate compared to a middle-aged candidate. We do not find evidence that people are ageist in penalizing younger or older candidates. This suggests that the current age profile of elected officials in government may have more to do with the structural factors that influence who runs and who wins, rather than the particular demands of voters.

To explore the robustness of this null finding, we also consider a second outcome tied to perceptions of the candidate’s ability to serve the district. We asked respondents to offer an evaluation of how well they thought the candidate would represent the interests of his constituents, if elected. The scale ranged from 0 (not at all) to 100 (very well). As shown in Fig. 1, we again fail to find evidence that people believe that a younger or older candidate will less successfully represent the interests of their constituents compared to a middle-aged candidate. Across two kinds of candidate ratings, people respond similarly to the younger, older, and middle-aged candidates.

We also consider the possibility of heterogeneous treatment effects by the age of respondents. If people prefer descriptive representation from those who share membership in their same generation, then younger people will feel more favorable toward a younger candidate and older people may respond more warmly to an older candidate. However, we fail to find evidence in support of our expectations. As shown in the supplemental appendix, the effects of the treatments do not significantly vary depending on the age of the respondent or the respondent’s generational cohort. Older participants and younger participants respond to the treatments in similar ways. This suggests that people do not have a strong preference for electing representatives in the same age cohort, perhaps because other attributes are more important than age in forming candidate evaluations.

We find little evidence that people prefer older candidates over younger candidates. People report similar confidence in the candidate’s ability to represent the interests of constituents regardless of the candidate’s age. As a final test, we consider the possibility that people form different impressions of the candidate as a function of politician age. It may be that the effects of age work through how people form impressions of the candidates and their traits, and the stereotypes they apply. As the last question in the experiment, respondents were given a list of nine traits and asked to check which ones they felt described the candidate they just read about. They could select as many or as few of the traits as they desired.

We first consider the effects of the treatments on perceptions of the competence of the candidate. We have reason to believe that younger politicians will be seen as less competent to serve in office as a result of their inexperience and immaturity, and that older candidates may be seen as less competent due to age stereotypes about declining mental performance. We include three indicators of perceived competence: whether the candidate was well-qualified to serve in the state legislature, knowledgeable about the issues, and inexperienced. Next, we consider indicators of perceived responsiveness, to see if the younger candidate is stereotyped as responsive and open-minded and older candidate is cast as disconnected and closed-minded. We consider two indicators of responsiveness, “cares about people like you” and “out of touch with ordinary people.” We next consider whether the age of the candidate affects whether people see the candidate as trustworthy and willing to stand up for what he believes. Finally, we include “liberal” and “conservative” to see if politician age is associated with perceptions of the candidate’s ideological leanings. Results are shown in Fig. 2.

Fig. 2
figure 2

Perceptions of Traits of Younger and Older Candidates, Experimental Results

Full size image

We find partial evidence that people make different inferences about candidate traits as a function of the candidate’s age. Considering first the indicators of perceived competence, the younger candidate is significantly less likely to be described as well-qualified to serve in office. In the control condition, 19% report that the candidate is well-qualified to serve, compared to only 13% of those reading about the younger candidate. We find that the older candidate also tends to be less likely to be seen as a well-qualified to serve in office (13%), though the difference between this condition and the control condition is only near significant (p < 0.06). When asked whether the candidate is knowledgeable about the issues, we find no differences across the three conditions. However, people are significantly more likely to describe the young candidate as inexperienced. 47% of respondents selected this as a term that described the younger candidate, versus 22% in the older candidate treatment and 23% in the control condition. Even though our participants are just as likely to say they would vote for the younger candidate, they nonetheless hold at least some reservations about the qualifications of a young person serving in government. These negative stereotypes could be damaging within campaigns, given that the public typically prefers candidates with greater expertise (Kirkland and Coppock 2018; Teele, Kalla, and Rosenbluth 2018).

When asked about whether the candidate is out of touch with ordinary people or cares about people like you, we find no statistically significant differences across conditions. This is consistent with our prior finding that people see the candidate as just as capable of representing the needs of their potential constituents regardless of the candidate’s age. Despite stereotypes that describe older people as rigid or out of touch, we fail to find evidence that people see older politicians as less responsive to their constituents. Likewise, we fail to find any differences in perceptions of the trustworthiness of the candidate across experimental conditions. When asked whether the candidate stands up for what he believes in, we find partial differences across experimental conditions. In the younger candidate condition, 40% say this describes the candidate, compared to only 33% in the control condition – a difference that falls just short of statistical significance (p < 0.07). In the case of the older candidate condition, 42% agree that the candidate stands up for what he believes in, which significantly greater than in the control condition. People perceive the force of a candidate’s convictions differently depending on the age of the candidate.

We also find partial evidence that people infer ideological leanings from the age of the candidate. In the control condition, 16% choose “conservative” as a term that describes the candidate, as do 15% who read about the older candidate. People who read about the younger candidate are significantly less likely to describe the candidate as conservative, with only 9% applying this term. We find that 10% choose the word “liberal” as one that describes the younger candidate, which is somewhat greater than is seen in the control condition (6%), but the difference here is just shy of statistical significance (p < 0.06). Among our participants, the age of the candidate is used to infer the candidate’s political leanings. Because millennials and members of Generation Z are typically more liberal in their political leanings than older generations (Parker and Igielnik 2020), age can be a heuristic that signals a candidate’s political leanings. We find additional evidence in support of this pattern when we consider a post-hoc test of heterogeneous treatment effects by respondent ideology, as shown in the supplemental appendix. We find that strong liberals voice higher levels of support for the younger candidate compared to the candidate in the control condition, while strong conservatives voice less support for the younger candidate relative the control condition.

To summarize, we find that the age of a candidate is not particularly important to people’s willingness to vote for that candidate. Young people are not uniquely drawn to younger candidates, and older people are as likely to support younger candidates as older ones. This is a good sign for young people interested in pursuing a political career, as age biases from the electorate will not present a major electoral hurdle. However, this also means that there may not be great demand to recruit more young people as candidates for office, as people seem as happy to endorse an older candidate as support a younger candidate. Younger candidates may mobilize young people to turn out to vote (Pomante and Schraufnagel 2015), but they do not seem any more likely to draw support among younger voters.

We find that younger candidates are likely to be confronted with negative stereotypes about their qualifications to serve in office. Reservations about the qualifications and experience of younger candidates may indirectly undercut people’s willingness to support younger candidates. We explored this possibility empirically using mediation analyses described in the supplemental appendix. We find partial evidence that the younger candidate treatment may have indirect effects of candidate support through two of the nine trait items: namely, perceptions of the candidate as inexperienced and less qualified to serve in office. This suggests that younger candidates can be electorally disadvantaged by the negative stereotypes people attach to their candidacy. While people are willing to support younger candidates, younger politicians may need to do more to convince voters of their competence to serve. Young conservative candidates may face the additional challenge of convincing their constituents to not conflate their age with their ideological commitments.

Survey analysis of politician age and approval of representatives in Congress

To what degree does a politician’s age matter once he or she is elected to office? We next consider the relationship between politician age and how people assess the way legislators represent their interests. We rely on survey data from Cooperative Election Studies from election years from 2006 to 2020. We focus on approval of members of Congress rather than vote choice, given that we do not expect age biases to be so potent as to interfere with people’s tendency to vote along party lines. While our experimental evidence informs whether people select candidates based on their age, our survey evidence allows us to explore whether age informs retrospective assessments of the job performance of elected representatives. We can move beyond asking people to assess hypothetical candidates to consider how people evaluate their own representatives and the impressions they have developed over time.

Our main focus is on whether the age of the representative in years is associated with higher or lower job approval, based on a question that asks people if they strongly approve, somewhat approve, somewhat disapprove, or strongly disapprove of the way their member of Congress is doing their job. Alongside our measure of legislator age, we include a number of control variables. We have reason to believe that the age of a politician may be associated with how they approach their job (e.g. Curry & Haydon 2018; Fengler, 1980; Schubert, 1988). As such, we include several controls to help better isolate the distinctive effects of age. The first is a measure of seniority, measured as the number of years since the representative was elected to this seat. Lawmakers who have served a longer tenure in their jobs have had more time to build a base of support in the constituency, so we wish to isolate the specific effects of age from the approval gains that can come from years of legislative service. We also control for legislative effectiveness, using scores created by Volden & Wiseman (2014). This measure draws on data on factors like bill sponsorship and committee action to assess how successful each representative is at shepherding policy ideas into law. We expect that lawmakers with higher levels of effectiveness will enjoy greater support from constituents.

We also include controls for shared partisanship and ideological congruence. We first include an indicator of whether respondents share the same party affiliation as their representative in Congress. We expect strong partisan effects, where people will be significantly more likely to approve of their representative when they share the same party allegiance. We also consider ideological similarity, with the expectation that representatives who cast votes consistent with the ideological preferences of constituents will enjoy higher approval. To test this, we include an interaction between a five-point measure of respondent ideology and the ideology of the representative as reflected in the votes they cast, measured as the first dimension DW-Nominate score (Lewis et al. 2021). We expect that increasing legislator liberalism will be associated with greater approval among self-identified liberals and greater disapproval among conservative respondents. We also include controls for other demographic differences, namely the race, ethnicity, and gender of representatives in Congress. At the individual level, we include demographic controls for age, education, gender, and race.Footnote 13

Because respondents are nested within districts, we use multilevel modelling and a random-intercept random-coefficient specification.Footnote 14 We report two models in Table 1. In the first, we consider the direct effects of a legislator’s age on approval. In the second, we consider the effects of representative age as conditional on the age of the respondent. In the first model, we find a significant negative effect of representative age on approval of members of Congress. Older legislators enjoy lower levels of job approval than their younger colleagues. The effect is substantively small. The predicted probability of strongly or somewhat approving of one’s representative is 60% when that representative is 30 years old, and drops to 58% when that representative is 80 years old, all else equal. In contrast to our experimental results that showed little preference or penalty associated with older candidates, we find here that older lawmakers enjoy lower levels of approval compared to younger representatives. And while our experiment showed that people tended to see younger candidates as less qualified and less experienced, we find no penalty associated with the youth of a representative in this data. Election to office may mute concerns about the qualifications of younger politicians.

Table 1 Age of Representative and Constituent Approval, 2006–2020 CES
Full size table

In the second model, we find that the penalties for legislator age are conditional on the age of the respondent. Younger respondents are more likely to give negative evaluations of an older representative than older respondents. In Fig. 3, we plot the marginal effect of representative age on approval over the range of respondent age. In Fig. 4, we plot the predicted levels of approval of members of Congress as a function of the age of the legislator for respondents aged 18, 48, and 78. For ease of interpretation, we combine probabilities of somewhat and strongly approving of the representative. We find that older respondents evaluate their member of Congress similarly whether she or he is younger or older. For those over age 55, the marginal effect of lawmaker age is positively signed, but statistically indistinguishable from zero. But for younger respondents, we find a significant negative marginal effect associated with legislator age, one that decreases in magnitude from early adulthood to middle age.

Fig. 3
figure 3

Marginal Effect of Representative Age on Approval, by Respondent Age, 2006–2020 CES

Full size image
Fig. 4
figure 4

Predicted Approval based on Respondent Age and Representative Age, 2006–2020 CES

Full size image

As shown in Fig. 4, a 78-year-old respondent evaluates a younger legislator similarly to an older representative, all else equal. However, a 48-year-old respondent has a 60% likelihood of approving of their representative when he or she is 30 years old, and only a 58% likelihood of approving of their representative if he or she is 80 years old. The penalty for representative age is even greater among younger respondents. For an 18-year-old in our sample, the predicted probability of approving of a 30-year-old representative is 60%. This drops to 54% in the case of a representative aged 80, an age penalty triple in magnitude to that applied by a 48-year-old respondent. While younger legislators are evaluated similarly by older and younger constituents, older lawmakers are less likely to secure the support of younger constituents.

The effects of legislator age are modest relative the importance of shared partisanship and ideology, which are our strongest predictors of representative approval. Those who share the same party affiliation as their member of Congress have a 76% likelihood of approving of their representative, while those who do not share the same party tie have only a 41% likelihood of approving.Footnote 15Footnote 16 People’s ratings of their representative also respond to the ideological tenor of lawmakers’ votes. Consider the representative with the most liberal voting history in 2020. A strong conservative in their district has a 26% likelihood of approving of their representative, while the most liberal respondents have an 89% likelihood of approval. The pattern reverses when considering ideological differences in predicted approval of the most conservative representative. Here, the most liberal respondents have a 16% likelihood of approving of their representative, while the most conservative respondents have an 87% probability of approving of their lawmaker. The effects of legislator age are quite small relative the effects of shared political outlook.

We find that the effects of legislator age are distinctive from the consequences of legislator seniority and effectiveness. Legislators who are more effective at their jobs do not enjoy significantly higher level of approval in our models. Greater seniority is associated with somewhat higher approval, where a lawmaker with a fifty-year tenure in office secures a predicted level of approval about three points higher than that of a freshman representative. We also find that the effects of the age of representatives on approval are often similar in magnitude to the effects of other demographic traits of lawmakers. We find about a two-point difference in approval of a 30-year-old representative compared to an 80-year-old representative. This is similar in magnitude in the penalty that Black and Latinx representatives face relative white members of Congress. In this dataset, Black representatives have a predicted level of approval three points lower than white lawmakers, and the predicted level of approval for Latinx representatives is about two points lower than that of white representatives.Footnote 17 Women in Congress earn lower levels of approval than men in this sample, but the effect size is small and less than that of age, at about one point difference.

We must acknowledge that with our observational data, we cannot know with certainty that this age penalty is solely due to the age of the representative, and not unmodeled aspects of job performance that might be correlated with member age but not captured in our controls for policy liberalism and time and performance in office. But given that we find that penalty for older age varies depending on the age of the constituent, we can say that older representatives – either because of their age or how they approach their job – are less likely to win the support of younger constituents in their districts.

Conclusions

The American electorate is graying. People are having fewer children, but enjoying longer life expectancy (Vespa, 2018). As older adults compose an increasing share of the electorate across decades to come, will that lead to greater demand for older politicians in state legislatures and in Congress? Our results suggest no. Even as seniors are mobilized by shared concerns like Social Security and protecting retirement benefits (Campbell, 2003; MacManus, 1995), we fail to find evidence that older people prefer age-based descriptive representation. Older people are not more likely to support older candidates over younger candidates. When evaluating job performance of incumbents, they offer similar ratings of their representative whether younger or older. Older people do not seem to prefer politicians who are closer in age to them.

To the extent to which we find evidence of distinctive age-based preferences among our respondents, it is among younger people. News accounts have argued that younger generations are frustrated with older politicians, but struggle to see ways to change the system (Astor, 2020; Taylor, 2018). Our results support this in part, as young people evaluate older representatives in Congress more negatively than younger representatives. Yet it is less clear that young people necessarily look to their peers in identifying paths to change. When candidate traits and campaign platforms are held constant, young people do not report a stronger preference for younger state legislative candidates than older ones. This is consistent with the patterns seen in the 2020 Democratic primary. Bernie Sanders was one of the oldest candidates in the primary race, but young voters were more likely to back him than younger candidates like Pete Buttigieg and Andrew Yang (Lardieri, 2020). It may be that young people want politicians who share the substantive concerns of their generation more so that just candidates who can descriptively represent them by age. While people may value descriptive representation from those who share the same group memberships (Mansbridge, 1999), we do not find much evidence that people are drawn to candidates who are closer in age.

Prior studies have suggested that a politician’s age may be more important than even attributes like the candidate’s race and gender in how people select who to vote for (Hainmueller, Hopkins, and Yamamoto 2014; Piliavin 1987; Sigelman & Sigelman, 1982). We do not find evidence that the age of a politician affects candidate support in our campaign experiment. In selecting candidates, people offer similar ratings of a young candidate, middle-aged candidate, and older candidate. However, when evaluating the job performance of elected representatives, age plays a somewhat greater role. We find that people are somewhat less likely to approve of older representatives than younger representatives. The effects are modest in magnitude, but often comparable to the effects of other demographic differences in legislator race, ethnicity, and gender for representative approval.

When thinking about why older politicians are over-represented in government, our results suggest this likely has more to do with structural factors like incumbency advantage than the biases of the electorate or public demand for older politicians. However, we cannot entirely rule out the possibility that voter biases lead to the under-representation of younger politicians. We find people hold negative stereotypes of the competence of younger candidates and perceive them as ideologically different than older candidates. These stereotypes may disadvantage the candidacies of younger politicians. Yet these negative stereotypes likely pose a hindrance rather than an insurmountable barrier to younger people seeking political office, as younger politicians seem to draw as much candidate support as older politicians in our experiment. While younger candidates may not always have as much political experience as older candidates, Hansen & Treul (2021) have argued that this may not always be a flaw in the eyes of voters. Candidates with less experience in politics can present themselves as being separate from the Washington establishment. Younger candidates are uniquely poised to capitalize on this framing, and are able to present themselves as political outsiders.