Theoretical Background

Unelected Representation

Constructivist theorists of political representation argue that political representation is not necessarily the result of elections (Dryzek and List Reference Dryzek and List2003; Saward Reference Saward2006; Disch Reference Disch2011). Where the standard account of political representation starts with elections and then asks whether the result is representative, constructivists start with acts of representation – representative claims – in which would-be representatives present themselves as champions of certain causes and groups, followed by acceptance or rejection by those witnessing these acts. Central to this constructivist turn in representation is the idea that representation should be viewed as a dynamic process of claim-making, an ongoing process of making and witnessing, accepting, and rejecting claims – in, between, and outside electoral cycles (Saward Reference Saward2010). Political representation is here defined as a two-way process in which both representatives and the represented play active roles in defining and shaping the terms of political representation.

Saward (Reference Saward2010) defines a representative claim as ‘a claim to represent or to know what represents the interests of someone or something’ (p. 38). To study these claims empirically, De Wilde Reference De Wilde2013 developed Representative Claims Analysis (RCA), which identifies three essential components of representative claims: the maker, the claim itself or policy demand, and the object or ascribed constituency. Representative claims can be found in various contexts, including social media, mass media, and parliamentary debates (Heinisch and Werner Reference Heinisch and Werner2019; Joschko and Glaser Reference Joschko and Glaser2019; Celis et al. Reference Celis, Childs, Kantola and Krook2014). These claims are made by both elected and unelected representatives and can encompass a wide range of policy demands on behalf of diverse groups (De Wilde Reference De Wilde2020), including transnational constituencies (Kinski and Crum Reference Kinski and Ben Crum2020). Typically, representative claims focus on broad interests or values rather than specific constituencies. Examples include statements such as ‘We demand justice’ or ‘The reform is needed to improve children’s education’ (Guasti and Geissel Reference Guasti and Geissel2019, 102). The public articulation of a political demand qualifies it as a claim (Koopmans and Statham Reference Koopmans and Statham1999), while the linguistic connection to a group or collective transforms it into a representative claim.

From a constructivist perspective, representation can, in principle, be ‘practiced, performed, claimed’ by anyone within and without the familiar institutions of electoral democracy (Saward Reference Saward2010). Specifically, this leads to an expansion of representation (Wolkenstein Reference Wolkenstein2021; Wolkenstein and Wratil Reference Wolkenstein and Wratil2020), as not only elected representatives can make representative claims, but also unelected representatives (Wolkenstein Reference Wolkenstein2021; Wolkenstein and Wratil Reference Wolkenstein and Wratil2020; Fossheim Reference Fossheim2022; Bovenkamp and Vollaard 2018).Footnote ¹ In conclusion, when representative claims are made by individual political activists who are not elected, they contribute to the universe of unelected representation (see Vik Reference Vik2025).

In short, we take a political communication perspective on the universe of unelected representation, demarcated by the occurrence of a particular subclass of speech acts in the public sphere. Unelected representatives, for the purpose of our study, are thus defined as political activists who formulate representative claims, that is, policy demands, on behalf of self-articulated constituencies in front of a wide audience. Unelected representatives are thus a specific subgroup of political activists, and they stand out through their performance of representative claims. In Figure 1, we have illustrated our definition exemplified by a claim from Greta Thunberg, the climate activist who has hit a political nerve with her activism and representative claims, mobilizing millions of young climate activists and becoming the Times Person of the Year in 2019 (Felsentahl Reference Felsentahl2019).

Figure 1. Illustrative Example of Unelected Representation: A Representative Claim by Greta Thunberg.

Unelected representation challenges conventional notions of political representation by bypassing key features of liberal democracy, with significant implications for its functioning. This shift emphasizes the importance of new, unconventional sites and practices of representation, pluralizing the concept and making it harder to understand through traditional political theory (Sintomer Reference Sintomer2013). Thus, current discussions on unelected representation focus largely on whether these representatives help or hinder democracy (Kuyper Reference Kuyper2016; Montanaro Reference Montanaro2012; Näsström Reference Näsström2015) and on what this expansionism means for democratic legitimacy (Schweber Reference Schweber2016; Disch, Sande and Urbinati Reference Disch, van de Sande and Urbinati2019; Montanaro Reference Montanaro2017; Disch Reference Disch2015; Kuyper Reference Kuyper2016; De Wilde Reference De Wilde2020). Without denying the remaining scope for normative assessments of the legitimacy of unelected representation, for the purpose of our study, our focus aligns with Saward Reference Saward2010 ideas on Weberian uses of legitimacy and constructivism; citizens ultimately determine the legitimacy of political representation – if they feel represented, the representative is legitimate.

Substantive Representation and Feeling Represented

According to Dahl Reference Dahl1971, substantive representation is essential for the functioning of a healthy democracy, as it allows representatives to use their positions of authority and influence to actively advocate for policies and decisions that will benefit their constituents (p. 1). This reflects the strong importance scholars of political representation and democracy, in general, have placed on substantive congruence – the principle that the policy preferences of elected officials should reflect the policy preferences of those they represent (Thomassen and Schmitt Reference Thomassen and Schmitt1999).Footnote ² Issue congruence is, therefore, by many considered to be an essential indicator of representation (Miller and Stokes Reference Miller and Stokes1963; Powell Reference Powell2004). In the circumplex model in political science as a theory of core affect, issue congruence is found to be affectively laden (Bakker, Schumacher and Rooduijn Reference Bakker, Schumacher and Rooduijn2021). Those with extreme attitudes experience more arousal in response to political rhetoric, while political rhetoric incongruent with prior attitudes evokes negative affect (Bakker, Schumacher and Rooduijn Reference Bakker, Schumacher and Rooduijn2021). This aligns with substantive representation in the way that we can explain stronger positive evaluations when the respondent views a claim that is congruent with their previously held beliefs, and the opposite effect when their views are incongruent with the message. Such findings imply that substantive congruence is conducive to people feeling represented, and thus, we formulate the following hypothesis.

The effectiveness of issue congruence can vary depending on the context, the salience of the issues, and the evaluation of the importance of the issues (Burstein Reference Burstein2003; Costello et al. Reference Costello, Toshkov, Bos and Krouwel2021; Page and Shapiro Reference Page and Shapiro1983). Saward (Reference Saward2006) argues that issue congruence may be less pronounced in situations where there is a high level of public trust in representatives, as constituents may be more willing to accept representatives who hold different views on certain issues. Thus, we will inspect how issue congruence matters in unelected representation on issues that vary in political polarization, namely climate action, immigration, and taxation (Pew Research Center 2019), and explore under which conditions issue congruence matters more or less.

Psychological Representation and Feeling Represented

The congruency model of political preference states that there are ‘interacting congruences among voters’ self-reported traits and values, voters’ perceptions of leaders’ personalities, politicians’ self-reported traits, and programmes of favoured political coalitions’ (Caprara and Zimbardo Reference Caprara and Zimbardo2004, 581). Similarly, the susceptibility to differences model explains that differences in personality can explain why communicated messages, in our case, representative claims, are appreciated by some and not by others (Valkenburg and Peter Reference Valkenburg and Peter2013). While so far largely overlooked in the literature of political representation, psychological congruence between representative claim-makers and members of the audience might thus be a key missing piece of the puzzle in understanding representative linkages. Personality is defined as ‘a multifaceted and enduring internal, or psychological, structure’ that is typically manifested as traits (Mondak et al. Reference Mondak, Hibbing, Canache, Seligson and Anderson2010). These traits are reflected as internal dimensions that predispose one to relatively stable patterns of experience and behaviour across situations (McCrae and Costa Reference McCrae and Costa2008).

Three of the most popular trait frameworks used in political science are the Big Five, HEXACO, and the Dark Triad. The Big Five consists of five traits: extraversion, agreeableness, conscientiousness, emotional stability (the inverse of neuroticism), and openness (Goldberg Reference Goldberg2013; McCrae and Costa Reference McCrae and Costa2008). HEXACO has the same traits as the Big Five but also includes a measure of honesty-humility (Lee and Ashton Reference Lee and Ashton2014). Meanwhile, the Dark Triad consists of three ‘socially malevolent’ traits: narcissism, psychopathy, and Machiavellianism (Jonason and Webster Reference Jonason and Webster2010). HEXACO’s honesty-humility measure is essentially the opposite equivalent to the common element shared by the Dark Triad variables (Lee and Ashton Reference Lee and Ashton2014). These traits have been found to be robust measurements, being relatively consistent across cultures, situations, and over time (McCrae and Costa Reference McCrae and Costa2008; Larsen and Buss Reference Larsen and Buss2010; Rogoza et al. 2021; Schmitt et al. Reference Schmitt, Allik, McCrae and Benet-Martínez2007).

Citizens tend to vote for politicians whose personality traits are in accordance with the ideology of their preferred political party, as well as politicians who possess their own personality traits (Caprara and Zimbardo Reference Caprara and Zimbardo2004). For example, citizens with dark traits prefer candidates with dark traits (Hart, Richardson and Tortoriello Reference Hart, Richardson and Tortoriello2018; Nai, Maier and Vranić Reference Nai, Maier and Vranić2021). In research that asked respondents to design their ideal politician, individuals sought to find leaders with traits that were congruent with their own (the congruency principle) but also those who possessed some key leadership qualities (the leadership appeal framework) (Aichholzer and Willmann Reference Aichholzer and Willmann2020). Furthermore, (Kalmoe, Gubler and Wood Reference Kalmoe, Gubler and Wood2018) tested the polarizing effects of common violent metaphors on issue preferences and found these effects to depend on participants’ aggressive personality traits. Similarly, Bakker, Schumaker and Rooduijn (Reference Bakker, Schumaker and Rooduijn2020) found that agreeableness had an independent effect on populist voting (irrespective of the ideological position of the populist parties), in which individuals who scored low on agreeableness were more likely to support populist parties. This effect has been found to be caused by the anti-establishment component of populist rhetoric, as an anti-establishment message is congruent with not being agreeable (Bakker, Schumaker and Rooduijn Reference Bakker, Schumaker and Rooduijn2020). In (political) marketing, psychological microtargeting is a hot topic, and several studies have shown that personality-tailored ads are very persuasive (Matz et al. Reference Matz, Kosinski, Nave and Stillwell2017; Zarouali et al. 2022). Together, these studies illustrate that congruence in personality matters for numerous processes related to political representation.

Aichholzer and Willmann (Reference Aichholzer and Willmann2020) argue that congruence in personality may foster ideological or value representation.Footnote ³ Certain personality traits can reflect values; for instance, agreeableness correlates positively with benevolence and traditional values, and these values are strong among populist supporters (Roccas et al. Reference Roccas, Sagiv, Schwartz and Knafo2002; Baro Reference Baro2022). Thus, it is possible that individuals who score low on agreeableness prefer politicians with the same trait, as this can be argued to be a form of value congruence. In addition to agreeableness, narcissism is a personality trait that is closely related to self-enhancement values (Kajonius, Persson and Jonason Reference Kajonius, Persson and Jonason2015). Agreeableness refers to the tendency to be likeable, agreeable, pleasant, and harmonious in relation to others (Gerber et al. Reference Gerber, Huber, Doherty and Dowling2012), and (grandiose) narcissism consists of grandiosity, a need for admiration, a lack of empathy, a sense of entitlement, and self-admiration (Rogoza et al. 2021). Both of these traits are what we consider to be value-associated traits, and we go into depth in our reasoning when selecting these traits in Appendix B1. We expect that the importance of personality-trait congruence, especially for the traits of agreeableness and narcissism, extends to unelected representation and that citizens feel more represented by activists who mirror their own personalities:

In addition to straightforward personality-trait congruence, there is also substantial evidence for personality-ideology congruence, as personality is strongly correlated with political preferences (Jost, Nosek and Gosling Reference Jost, Nosek and Gosling2008; Fatke Reference Fatke2017; Ekstrom and Federico Reference Ekstrom and Federico2019; Jonason Reference Jonason2014; Aichholzer, Danner and Rammstedt Reference Aichholzer, Danner and Rammstedt2018; Mondak and Halperin Reference Mondak and Halperin2008). From the Big Five framework, two have emerged as robust predictors of political preferences: openness and conscientiousness. Openness is characterized by the willingness to try new experiences and the interest in abstract ideas and concepts, while conscientiousness is characterized by a sense of responsibility, organization, and self-discipline (Costa, McCrae and Dye Reference Costa, McCrae and Dye1991). Individuals who score high on openness tend to prefer left-wing parties and ideologies (for example, political liberal policies), whereas individuals scoring high on conscientiousness tend to prefer parties and ideologies from the right (for example, political conservative policies), and this effect is found in most Western countries and is stable over time (Jost, Nosek and Gosling Reference Jost, Nosek and Gosling2008; Fatke Reference Fatke2017; Ekstrom and Federico Reference Ekstrom and Federico2019). This can be argued for by the fact that openness has clear implications for how people tolerate uncertainty and ambiguity. This is more prevalent in left-leaning ideologies seeking social change (Ekstrom and Federico Reference Ekstrom and Federico2019). Conversely, right-leaning parties/politicians/ideologies promote stability, duty, and obedience, which likely resonates with individuals scoring high on conscientiousness; however, this relationship has more mixed and somewhat weaker evidence (Ekstrom and Federico Reference Ekstrom and Federico2019). Together, these studies provide evidence of the relationship between personality and ideology. Therefore, we expect that when the ideological position of the representative claim matches the respondent’s personality traits, respondents feel more represented.

There is some overlap between psychological and descriptive routes to representation since both represent dimensions of similarity between representatives and constituents. Yet descriptive representation has typically been studied with a focus on demographic characteristics, which are typically externally identifiable and related to people’s position in society. By contrast, personality traits are internal, relatively stable psychological characteristics that can be reflected in thoughts, communication, and behaviour. Psychological representation is, therefore, also different from substantive representation, which is traditionally oriented toward policy advocacy and, hence, an external act of responsiveness to the interests of the represented (for example, Pitkin Reference Pitkin1967, 209; see also Lavi et al. Reference Lavi, Treger, Rivlin-Angert, Sheafer, Waismel-Manor, Shenhav, Harsgor and Shamir2024). In Pitkinian terms, our pathways resonate with the descriptive, substantive, and symbolic dimensions of dyadic representation (Lavi et al. Reference Lavi, Treger, Rivlin-Angert, Sheafer, Waismel-Manor, Shenhav, Harsgor and Shamir2024; Pitkin Reference Pitkin1967), albeit that our understanding of psychological representation focuses on persons and their performative style and elaborates its psychological underpinnings much more elaborately than most understandings of the symbolic dimension. As part of psychological representation, it also relates to personality-ideology congruence, as hypothesized in H4, where people with certain personality traits are drawn towards certain ideological orientations; this pathway could indirectly contribute to substantive representation. It goes beyond the scope of the present paper to investigate how these routes interact with and complement each other.

Measures

The full survey design consisted of a pretreatment questionnaire, the conjoint module, a vignette experiment for another study, and a posttreatment questionnaire. A detailed account of the measures and their justifications can be found in Appendices B1 and B2.

In the pretreatment questionnaire, we asked respondents questions relating to our three types of representation. To establish (1) descriptive representation, we asked respondents about their (a) gender and (b) age. Ethnicity, for instance, becomes practically infeasible as our design is conducted in several European nations with different ethnic majorities/minorities, which would lead to difficulties in maintaining experimental control and statistical power. To measure (2) substantive representation, we asked respondents to indicate their pro or anti stance on our three issues on a ten-point scale where 1 meant that they ‘fully agreed’ with the statement on the left, and 10 meant that they ‘fully agreed’ with the statement on the right: (a) issue positions on climate protection were measured with a scale ranging from ‘climate protection should take priority, even at the cost of economic growth’ to ‘economic growth should take priority, even at the cost of climate protection’; (b) for issue positions on immigration, the endpoints of the scale consisted of the statements ‘immigration strengthens our society’ and ‘immigration is a threat to our society’ and; (c) for issue positions on taxation, we pitted the statement ‘taxes should be raised to increase public services’ against the statement ‘taxes should be cut, even at the expense of cutting public services’. For psychological representation (3), we measured (a) narcissism ( $\alpha $ = 0.82) using the four items from the ‘Dirty Dozen’ (Jonason and Webster Reference Jonason and Webster2010), (b) agreeableness ( $\alpha $ = 0.69), (c) openness ( $\alpha $ = 0.61), and (d) conscientiousness ( $\alpha $ = 0.63) were measured by the four-item scales from the Mini IPIP (Donnellan et al. Reference Donnellan, Oswald, Baird and Lucas2006).

Congruencies

For all the routes to representation, we created measures of congruence by contrasting the pretreatment measurements with the attributes illustrated in the conjoint design. In other words, we consider congruence to be a match between respondents’ self-reported attributes with the stimuli presented in the conjoint. This process was inspired by how previous work on issue congruence was coded (Bakker, Schumacher and Rooduijn Reference Bakker, Schumacher and Rooduijn2021) and was specified in our pre-registration. We explain the coding briefly below, and in Appendix B3, we have illustrated the coding process.

For demographic congruence, we create congruence indexes for gender and age. Gender congruence is simply whether the gender of the respondents matches with the displayed gender of the claimant (0 = incongruent, 1 = congruent). If respondents reported other or nonbinary as their gender, this was coded as incongruent. For age congruence, we created a continuous scale of congruence by estimating the age difference between the respondent’s age (18–90) and the activist’s (GenZ: 20, Millenial: 35, GenX: 55, Boomer: 70), and reverse-coded the index so that a higher score represented high levels of congruence. To illustrate, when a 20-year-old respondent witnesses a 20-year-old claimant, they are given a score of 70 (maximum congruence), whereas when a 90-year-old respondent witnesses the same activist, their age congruence score is 0 (minimum congruence).

For issue congruence, we computed degrees of congruence with each issue by contrasting the issue position of the respondent with the policy demand of the claim displayed in the conjoint design (−4 = maximum degree of incongruence, 4 = maximum degree of congruence). A positive score ( $0 \lt x \le 4$ ) means that the policy demand in the representative claim was congruent with the respondent’s issue position (for example, anti-immigration position and anti-immigration claim). Negative scores ( $ - 4 \ge x \gt 0$ ) indicate incongruence between the respondent’s issue position and the policy demand portrayed in the claim (for example, anti-immigration attitude and pro-immigration claim). Again, we assume that the claim is more congruent for participants with more extreme attitudes, as done in Bakker, Schumacher and Rooduijn Reference Bakker, Schumacher and Rooduijn2021. A score of ‘0’ indicates that the respondent has a neutral issue position, meaning that none of the policy demands in the claim are congruent or incongruent with the respondent’s issue position.

For personality congruence, we created variables indicating (1) trait/trait congruence, the match between the traits indicated in each profile and respondents’ self-reported trait; and (2) trait/ideology congruence: the match with the respondent’s self-reported traits (openness and conscientiousness) and the ideological position of the claim (left-leaning claims and right-leaning claims). We calculate the congruence in a similar manner to issue congruence. We computed degrees of personality trait congruence by contrasting the personality trait displayed in the conjoint design with the respondent’s score of the personality trait (−2 = maximum degree of incongruence, 2 = maximum degree of congruence). A positive score ( $0 \lt x \le 2$ ) means that the trait manipulated in the conjoint was congruent with the respondent’s score on that trait (for example, a respondent with low agreeableness witnessed a claimant with low agreeableness). For personality-ideology congruence, we created a congruence index by contrasting the ideological position of the claim (left-leaning; pro-climate, immigration, taxation or right-leaning; anti-climate, immigration, and taxation) with the respondents’ scores of openness and conscientiousness (−2 = maximum degree of incongruence, 2 = maximum degree of congruence). A positive score ( $0 \lt x \le 2$ ) means that the ideological position in the conjoint design was congruent with the respondent’s score on either openness or conscientiousness (for example, a respondent with high openness witnessed a left-leaning claim).

To make the continuous congruence measures more comparable and interpretable while retaining the fine-grained information from the continuous measure, we rescale them to range from 0 (minimum congruence) to 1 (maximum congruence).Footnote ⁷ In our pre-registration we did not specify that for H2 (issue congruence) and H3 (personality congruence), we will have missing values for the subcongruencies we test in the hypotheses, that is congruence for the three issues (a) climate, (b) immigration, and (c) taxation and for the personality traits (a) narcissism and (b) agreeableness. When the respondent, for instance, witnessed a pro-climate claim, we cannot calculate immigration congruence. Due to our high statistical power, we code them as missing values (NAs) to test the pre-registered hypotheses. In our exploratory analysis, in which we aim to compare the routes to representation, we take into account higher-order congruence for H2 and H3. In other words, we look at issue congruence and personality congruence overall, as it was visible in the conjoint design. This is illustrated in Appendix B3.

Controls

Although a fully randomized design is expected to eliminate the impact of confounding variables, including controls for potentially influential factors that may vary across the countries and influence our dependent variable in our analysis, can help prevent any unidentified confounding effects that may arise. Therefore, we specified controls a priori as social media use, political interest, and social trust (see Appendix B2 for the exact wording of all measures).Footnote ⁸ In our sample descriptives, we see that these variables do differ substantially between the countries (see Appendix A2).

Descriptive Representation

As seen in Table 1, consistent with H1, both gender congruence ( $\beta $ = 0.05, SE = 0.02, p $ \lt $ .003) and age congruence ( $\beta $ = 0.25, SE = 0.03, p $ \lt $ .001) are positively associated with feeling represented. When respondents witnessed a profile that was gender-congruent, holding other variables constant, they had a 0.05 increase in the feeling represented scale, a slightly higher feeling of representation compared to respondents who witnessed a gender-incongruent profile. When the age congruence goes from a maximum degree of incongruency to a maximum degree of congruence, there is an average increase of 0.25 in the feeling represented measure. As the dependent variable is measured on a six-point scale, the substantive size of both effects is relatively low. Therefore, sharing demographics such as gender and age makes citizens feel slightly more represented, which supports both H1a and H1b.

Substantive Representation

We find large and positive effects on feeling represented when the demands in the representative claims are congruent with the issue positions of the participant on climate protection ( $\beta $ = 1.92, SE = 0.05, p $ \lt $ .001) and immigration ( $\beta $ = 2.15, SE = 0.05, p $ \lt $ .003), but an insignificant and slightly negative effect for taxation ( $\beta $ = 0.05, SE = 0.06, p $ = $ .33). It is interesting to note that the effect of issue congruence varies so much depending on the issue. Immigration has the largest effect, eliciting an average increase of 2.13 on the six-point feeling represented measure as we move from the maximum degree of incongruency to the maximum degree of congruence. Congruence on climate protection has a slightly smaller effect, but an average increase of 1.92 in the dependent variable still represents a large effect. In comparison, congruence on taxation leads to a statistically non-significant decrease of 0.05 in people’s feeling represented level. Thus, H2a and H2b are supported: seeing a claim that is congruent with your position on climate protection and immigration increases your level of feeling represented both substantially and significantly. Surprisingly, witnessing a claim that is congruent with one’s position on taxation has no effect on feeling represented. Therefore, no support is found for H2c. One possible explanation for this finding could be that prior research has found that the representation gap between voters and political candidates on economic issues is smaller than the representation gap between on cultural issues (Dalton Reference Dalton2017). From this perspective, the representation gap that can be filled in by claim-making from unelected representatives might be larger for cultural issues than for economic issues. Still, the finding could also be due to our operationalization of the pro- and anti-tax claims in the conjoint, which specified taxing the rich. While there was a substantial share of respondents who agreed with the statement against higher taxes prior to the treatment (see Appendix A2), we observed a strong preference among many respondents in favour of the pro-tax message (see Appendix C1 and Appendix C3.4). In our subgroup analysis of descriptive marginal means (available in Appendix C3.5), we discovered that responull findingndents who were pro-taxation felt mostly represented by the pro-taxation claim and not by the anti-taxation claim. Interestingly, even the respondents who reported being anti-taxation preferred the pro-taxation statement and did not even feel represented by the anti-taxation statement, which might explain this unexpected null-finding. We return to this point in the concluding discussion.

Psychological Representation

The results of our analysis reveal that both narcissism congruence ( $ \beta $ = 0.25, SE = 0.04, p $ \lt $ .001) and agreeableness congruence ( $\beta $ = 0.44, SE = 0.04, p $ \lt $ .001) are positively associated with higher levels of feeling represented, as expected. These effects are considered small to moderate. These results suggest that individuals who witness unelected representatives with similar traits tend to feel more represented, and we conclude that the evidence supports H3a and H3b.

Our analysis of the relationship between personality-ideology congruence and feeling represented yielded mixed results. Our regression analysis revealed that openness-ideology congruence, specifically a congruence between liberal representative claims and respondents’ levels of openness, is positively associated with feelings of representation ( $ \beta $ = 1.239, SE = 0.05, p $ \lt $ .001). This supports H4a. However, we found the opposite effect for personality-ideology congruence involving conscientiousness. The results indicate that when the ideological position of the representative claim is matched with respondents’ levels of conscientiousness, it leads to a significant and substantial decrease in feelings of representation ( $\beta $ = -1.015, SE = 0.04, p $ \lt $ .001). Therefore, hypothesis H4b is rejected. These effects could be due to the fact that the pro-statements resulted in respondents feeling the most represented (see Appendix C1 and C3.5).

Exploring the Role of Contextual and Individual Factors

We conclude the analyses by estimating a model with the congruencies at the conjoint levels (gender, age, issue, personality-trait, and personality-ideology) to compare the routes to representation and explore how contextual and individual factors influence the way the routes to representation work. Again, we use OLS regression with clustered standard errors. Importantly, our results are confirmed when controlling for other congruencies; gender congruence ( $\beta $ = 0.051, SE = 0.01, p $ \lt $ .001), age congruence ( $\beta $ = 0.24, SE = 0.03, p $ \lt $ .001), issue congruence ( $\beta $ = 1.45, SE = 0.03, p $ \lt $ .001), personality-trait congruence ( $\beta $ = 0.35, SE = 0.03, p $ \lt $ .001), and openness-ideology congruence ( $\beta $ = 0.77, SE = 0.05, p $ \lt $ .001) predict increased levels of feeling represented, and conscientiousness-ideology congruence ( $\beta $ = −0.325, SE = 0.04, p $ \lt $ .001) predicts decreased levels of feeling represented (see Appendix C2 for the model with all congruencies). As robustness checks, the manner in which binary coded congruencies influence feeling represented was investigated with two estimations: (1) Average Marginal Causal Effect (AMCE) and (2) descriptive marginal means. These are illustrated in Appendix C3.5 and largely replicate results from the OLS model in Appendix C2.

In the next step, we perform exploratory analyses to investigate potential contextual and individual factors that may influence how the routes to representation operate.Footnote ⁹ Specifically, we perform subgroup analyses by (a) gender, (b) age category, (c) political interest, and (d) country. The results are illustrated in Figure 3

Figure 3. Coefficient plots based on OLS models: subgroup analysis by (a) gender, (b) age category, (c) political interest, and (d) country. Standard errors are clustered at the respondent level.

From these plots, we see that overall issue congruence matters most for feeling represented ( $\beta $ = 1.45, SE = 0.03, p $ \lt $ .001). Gender congruence has a larger and significant effect for men ( $\beta $ = 0.07, SE = 0.02, p $ \lt $ .003) and a smaller significant effect for women ( $\beta $ = 0.04, SE = 0.02, p $ \lt $ .05). Age congruence also has a slightly larger effect for men ( $\beta $ = 0.32, SE = 0.04, p $ \lt $ .001), whereas the openness-ideology congruence has a larger effect for women ( $\beta $ = 0.96, SE = 0.07, p $ \lt $ .001). Baby boomers are more influenced by issue congruence ( $\beta $ = 1.96, SE = 0.06, p $ \lt $ .001), and gender congruence has a stronger effect for members of Generation Z ( $\beta $ = 0.17, SE = 0.04, p $ \lt $ .001). Having low political interest diminishes the effect of issue congruence ( $\beta $ = 1.00, SE = 0.05, p $ \lt $ .001). We also discover country differences, as in Romania ( $\beta $ = 0.38, SE = 0.07, p $ \lt $ .001), issue congruence has smaller effects on feeling represented compared to Sweden ( $\beta $ = 2.10, SE = 0.06, p $ \lt $ .001). Gender congruence, in contrast, has larger effects in Romania ( $\beta $ = 0.16, SE = 0.03, p $ \lt $ .001), with insignificant coefficients in all other countries of analysis. Overall, the effect of personality-trait congruence seems to be most stable across countries. The full regression table for all subgroup analyses is available in Appendix C4. In conclusion, we have strong evidence for the relevance of the different routes to representation, but we also discovered that the routes are not always straightforward.

Descriptive Representation

Our findings indicate that demographic congruence also matters in unelected representation. Gender and age congruence predicted higher levels of feeling represented, although the effect sizes are small, especially for gender congruence. We highlight that the effect holds up when controlling for other congruencies, that small effect sizes in such social science experiments are common (Funder and Ozer Reference Funder and Ozer2019), and that the gender cue was only given in terms of a picture and name. Making the gender cue more salient (for example, by including statements like ‘as a man, I believe in climate change’ or ‘as a man, I am fighting for change’), could potentially have made the gender congruence effect stronger. Additionally, although gender congruence had almost no effect in Sweden, it had effect sizes comparable to issue congruence in Romania; the effects were also larger for younger respondents and larger for men. This poses important questions related to internalized gender scripts, discussions of quotas, and the representativeness of both formal and informal political spaces.

First, although women have been found to report that they want or need women representatives (Arnesen and Peters Reference Arnesen and Peters2018), actual preference for gender in leadership could differ due to internalized gender expectations and gender stereotypes. For example, previous research has shown that both men and women perceive men as better political leaders (Bos et al. Reference Bos, Greenlee, Holman, Oxley and Lay2022). In addition, there is a growing body of evidence suggesting that objective and subjective marginalization do not necessarily coincide, and middle-aged white men report feeling more marginalized despite being objectively well-represented in politics descriptively (Bollwerk, Schlipphak and Back Reference Bollwerk, Schlipphak and Back2022). Is it possible that descriptive representation, therefore, is more important for individuals who feel subjectively marginalized due to these demographic traits rather than being objectively politically marginalized? And how does such a relationship work, taking into account gendered political socialization? Such interactions are important to explore in the debate on the meaning of descriptive representation.

Second, our results point to the relevance of descriptive representation in the public sphere. If prominent voices in the informal public sphere constituted by mass media, social media, and ‘the streets’ do not reflect society at large, this may lead to unbalanced feelings of representation among the population. This has implications for media organizations, such as TV channels and newspapers, when they decide whom to feature as talk show hosts, pundits, and columnists. It has implications for society at large when discussing how to deal with the online harassment and trolling of women, minorities, and marginalized groups, which could lead influential unelected representatives to retreat from the public sphere.

Third, our findings raise the question of whether unelected representation could potentially offset a lack of descriptive representation in formal arenas. We know that women, young people, and minorities are classically underrepresented in legislatures and other formal political bodies (Ruedin 2015). Can the recent prominence of young female activists in the public sphere, from Greta Thunberg to Malala Yousafzai, compensate for this in terms of making citizens feel represented, nevertheless? These are questions that clearly go beyond the scope of this paper, but our findings invite such questions.

Substantive Representation

Substantive representation turns out to be the ‘road most travelled’ towards representation in our study – having overall the largest impact on our respondents’ degree of feeling represented. Our design allows us to inspect issue congruence closely and explore meandering paths that may occur toward substantive representation. Congruence in preferences on taxation, for example, seems to have little effect on feeling represented. A majority reported preferring the pro-tax message, but they reported being more conservative in their actual tax positions. This could be due to the nature of the stimuli, as the stimuli in the conjoint design specifically mentioned taxing the rich and arguing that taxation can combat inequality, whereas the issue positions were more on raising or lowering taxes for public services. This also highlights something important with our design: the potential priming effects are likely to be stronger for issue congruence, given the similarity between the issue positions in the pretreatment questionnaire and those presented in the stimuli, especially for climate and immigration. An alternative or additional explanation is that in highly polarized issues and in issues with high salience (Pew Research Center 2019), such as climate change and immigration, substantive representation may be more important. Economic dimensions are known to have larger representation gaps in the elected arena (Dalton Reference Dalton2021), so this could be a substantive finding. Taxation may also be a less important topic in activism than the other two issues, at least in the European context. Either way, our dual findings that issue congruence on climate change and immigration have major effects on citizens’ feelings of being represented, while issue congruence on taxation does not raise follow-up questions on why this is the case.

Furthermore, the impact of substantive representation varies between countries; it seems to matter less in Romania, which has a lower quality of government, more among Baby Boomers, and more among politically interested individuals. This resonates with general biases in survey research: (i) self-selection bias: the average citizen is not very politically interested, but panel respondents in political science often are, and (ii) Western bias: much work is done in Western, rich, and affluent countries and effects may be culturally sensitive and might not extend to other contexts (Henrich, Heine and Norenzayan Reference Henrich, Heine and Norenzayan2010). Among groups that are often neglected in empirical political science, representation – and other political processes – could work very differently, and future research should aim to dissect how feeling represented through representative claims by both unelected and elected representatives works in these contexts. Overall, our paper shows that substantive representation – although the biggest driver for feeling represented – has important contextual and individual limitations.

Psychological Representation

We find support for the functioning of personality-trait congruence as a route to representation, albeit with small effect sizes. The level of representation felt by respondents increased when the personality traits signalled by unelected representatives in our stimuli, such as narcissism and agreeableness, matched those of the respondents. It is essential to consider the conjoint design when interpreting these results. The innovative approach of manipulating personality traits within brief formats, while common in marketing and psychological tailoring research (Matz et al. Reference Matz, Kosinski, Nave and Stillwell2017), differs from the longer text descriptions typically used in political science studies (Laustsen Reference Laustsen2017; Nai, Maier and Vranić Reference Nai, Maier and Vranić2021). The brevity required by our conjoint design may have constrained our findings on personality-trait congruence, as manipulations of gender, age, and issue positions can be inferred more directly from short text descriptions. Consequently, the modest effect of personality-trait congruence on feelings of representation might be more pronounced in real-life scenarios, where leaders consistently express their personality traits in more substantial ways. It could be that there are other manipulations that would be more effective in increasing feelings of representation, and we see it as a fruitful avenue for personality research to develop more evidence-based personality stimuli and assess the effectiveness of shorter text stimuli.

Relatedly, it will also be relevant for future research to investigate how personality congruences beyond agreeableness and narcissism influence experiences of political representation. Moreover, the reasoning behind the meaningfulness of trait congruence needs to be further investigated. As highlighted by Aichholzer and Willmann (Reference Aichholzer and Willmann2020), this congruence finding is likely due to some value-related representation, in which we may deem those similar to us, in terms of traits, to possess similar values, which could influence leadership appeal. In a time when psychological microtargeting is becoming increasingly common and has persuasive appeals (Zarouali et al. 2022), we need to dissect the meaningfulness of personality trait congruence.

We find mixed support for the congruence between the ideological position of the claim and respondents’ personality traits (personality-ideology congruence). Openness worked as hypothesized: when the claim was progressive (pro-immigration, climate, and taxation) and the respondents scored high on openness, the levels of feeling represented were substantially higher. In contrast to our expectations, when the claim was conservative (anti-immigration, climate, and taxation) and the respondent scored high on conscientiousness, the levels of feeling represented decreased substantially. The evidence for conscientiousness and preferences for conservative ideology is slightly more mixed or weaker (Ekstrom and Federico Reference Ekstrom and Federico2019), but it is still surprising to find a significant effect in the opposite direction. This is likely due to the strong preference in favour of the progressive positions (pro-climate action, immigration, and taxation) among our respondents, despite our sample being relatively equal in terms of their political ideology. This could be due to undiscovered confounding factors that influence the persuasiveness of the policy demand in the claim. In any case, our findings raise important questions regarding the relationship between personality and ideological positions.

Our inclusion of psychological factors is relatively novel in the study of political representation, certainly in comparison to descriptive and substantive factors in representation, which are well-established in the literature. Our findings provide evidence that psychological mechanisms, in terms of both personality-trait congruence and personality-ideology congruence, matter for feeling represented. We hope that further studies, both theoretical and empirical, will investigate the role psychology plays in political representation. Our results also open up avenues for future work on the psychological foundations of substantive representation. For example, the relationship between personality and ideology is complex, and in recent research, there is evidence that the relationship between them is bidirectional (Bakker, Lelkes and Malka Reference Bakker, Lelkes and Malka2021). Thus, investigating how personality may drive issue positions and how these dynamics contribute to substantive representation could be an important new avenue in the study of political representation.

Continuing on the Route Towards (Unelected) Representation

This paper presented and mapped out several routes to (unelected) representation. The psychological emphasis on the importance of homophily seems to travel to political representation; citizens feel more represented through claims made by unelected representatives that match them in terms of gender, age, issue positions, and personality traits, as well as through issue positions that match their personality traits. The possible intersections of these routes remain to be studied. This was beyond the scope of the present paper, but future research should aim to understand how the routes interact with each other, and it should explore the psychological foundations that may drive substantive and descriptive representation.

Further research could also investigate what people think of or consider when asked whether they feel represented by claims made by unelected representatives. The operationalization of our main dependent variable corresponds to that of other studies on political representation (Kaplan et al. Reference Kaplan, Sheafer, Waismel-Manor and Shenhav2023; Dvir-Gvirsman et al. Reference Dvir-Gvirsman, Tsuriel, Sheafer, Shenhav, Zoizner, Lavi, Shamir and Waismel-Manor2022; Duell et al. Reference Duell, Kaftan, Proksch, Slapin and Wratil2023) and the fact that routes to representation we know to function in the electoral arena are also found to empower feelings of representation by unelected representatives in our study supports the validity of our measurement. But since ours is the first study to apply this measurement outside the electoral arena, it is nevertheless worthwhile to investigate more in-depth what it means when people say they feel represented by an unelected representative making a representative claim.

In our study, we also manipulated a specific form of representative claims: interest/value claims (Guasti and Geissel Reference Guasti and Geissel2019). These claims do not include an explicit linkage with the self-articulated constituency but rather invoke a generic ‘we’ or ‘us’ on whose behalf the claimant claims to speak, allowing anyone to associate themselves with this constructed collective. We encourage future research to claim, with an explicit linkage to various specific constituencies, such as Greta Thunberg’s ‘future generations’ or Bono’s ‘Africa’. Experimental designs could, for instance, differentiate between the different types of claims as postulated in Guasti and Geissel Reference Guasti and Geissel2019 and compare effects on feelings of representation.

Further research also needs to address the democratic and normative implications of feeling represented via unelected representation. Unelected representation might be a supporting factor to the legitimacy of electoral representative democracy, or it may be detrimental (Kuyper Reference Kuyper2016; DeWilde Reference De Wilde2020). We have focused here on whether citizens feel represented through unelected representation and how (through which pathways). We have not asked whether citizens feel strongly or sufficiently represented by unelected representatives, implying that there would no longer be a need for elected representatives, nor have we compared feelings of representation as a result of activists elevated to the status of unelected representatives through successful claim-making to elected representatives, or to more formal unelected representatives, like appointed civil servants (for example, central bankers or key figures in international organizations), trade union leaders, or religious leaders. We therefore strongly encourage future research to investigate the implications of feelings of representation by unelected representatives for the legitimacy of electoral representative democracy. How do citizens feel about key institutions of representative democracy and their elected representatives when they feel represented by unelected representatives? Which factors might play a role in determining whether unelected representation supports or undermines the legitimacy of electoral democracy? The findings of this study present an invitation to pursue this line of inquiry, not a final answer.