A burgeoning line of party research finds evidence over time and across countries that parties rely extensively on group appeals (Dolinsky Reference Dolinsky2023; Horn et al. Reference Horn, Kevins, Jensen and Van Kersbergen2021; Huber Reference Huber2022; Huber and Haselmayer Reference Huber and Haselmayer2025; Riethmüller and Franzmann 2025; Thau Reference Thau2018, Reference Thau2019). Such group appeals are thought of as a strategy allowing parties and their candidates to connect to socio-demographic groups and build a reputation as a defender of the interests of specific groups. In this way, group appeals can polarize voters and contribute to well-documented social divides in voting behavior (Bornschier et al. Reference Bornschier, Häusermann, Zollinger and Colombo2021; Lipset and Rokkan Reference Lipset and Rokkan1967). Recent work on voters’ responses to group appeals provides observational evidence that group appeals correlate with electoral support among the groups that parties appeal to (Stuckelberger and Tresch Reference Stuckelberger and Tresch2024; Thau Reference Thau2021). In addition, experimental evidence points out that the effectiveness of symbolic group appeals – that is, appeals that do not refer to material or policy-based interests (Thau Reference Thau2021) – is similar to that of policy-centered appeals to specific groups (Finseraas et al. Reference Finseraas, Heath, Langsæther and Smeets2025; Robison et al. 2021).
What we know about how these policy-less, symbolic group appeals affect voter support remains limited, however. While there is a long-standing interest in how political elites attract social groups, existing research tends to focus on just one group in each study, such as women (Kam et al. Reference Kam, Archer and Geer2017), Latinos (Alamillo and Collingwood Reference Alamillo and Collingwood2017), or Christians (Margolis Reference Margolis2018). Moreover, most studies use their own unique conceptions of group appeals, which limits the inferences that can be drawn about the effectiveness of group appeals across studies. Robison et al. (2021) and Haffert et al. (Reference Haffert, Palmtag and Schraff2024) do examine voter effects of group appeals, as understood in the party research cited above. However, their evidence concerns only class appeals (Robison et al. 2021) or only rural/urban appeals (Haffert et al. Reference Haffert, Palmtag and Schraff2024). The first objective in this article is therefore to evaluate whether symbolic group appeals shape candidate support beyond a narrow focus on one particular group. Specifically, the effects of appeals based on class, place (rural or urban), education, age, gender, and ethnicity (minority or majority) are examined.
We also lack insights about the conditions under which group appeals have more or less of an impact on voters’ preferences. Our second objective is therefore to theorize and test, across a range of social categories, which factors strengthen citizens’ responsiveness to group appeals. Building on the social identity and reference group literatures, we argue that the strength of citizens’ social identities and their perceptions of the deservingness of specific groups are key moderators of the effect of group appeals. Examining these moderators of the effect of group appeals is important for our understanding of such effects. Thus, we need to investigate the conditions under which the appeals have stronger or weaker effects in order to understand the underlying mechanism. In this way, we also provide a framework for reconciling current contradictory findings in the literature (compare, e.g., Robison et al. 2021 and Hersh and Schaffner Reference Hersh and Schaffner2013).
We rely on experimental designs embedded in two surveys among British voters.Footnote 1 In Study 1, we included vignettes with appeals based on class, place, education, and age (N=1,500). The results of this study indicate that the effect of group appeals is not limited to appeals based on class. We also find that there is important variation between the appeals in terms of how both the in- and the out-groups respond to these appeals. These results hint at the presence of conditioning factors, which we examine in our second, preregistered study. Study 2 consists of a series of vignette experiments that were again fielded among a sample of British respondents (N=3,200). Study 2 aims to corroborate Study 1 but also extends it by including a larger number of appeals and, crucially, measures of two hypothesized moderators: social identity strength and perceptions of deservingness. The results of Study 2 also show that group appeals increase candidate support across a range of different types of groups. In terms of the conditions to this effect, we find that the strength of individuals’ social identities somewhat moderates the effects of group appeals. However, deservingness perceptions appear to be more important. We find strong evidence that the degree to which the group that is appealed to is perceived as deserving better life conditions shapes how the candidate making the appeal is evaluated.
Group appeals and electoral support
Work that has studied the electoral impact of social divides has described and explained their effect as a process that is not exclusively bottom-up (Enyedi Reference Enyedi2008; Ford and Jennings Reference Ford and Jennings2020) but partly results from the mobilization of group differences. Parties are important actors in this process of mobilization, and they can rely on different strategies. For example, they can present candidates that have particular characteristics or specific social backgrounds as a way to provide a cue to groups that share these characteristics (Barreto Reference Barreto2007; Dolan Reference Dolan2008). The effectiveness of such a strategy is evident from the work of Heath (Reference Heath2015) for the British case. Parties can also clarify their intention to represent a specific group by means of the issue and policy positions they take. By taking an economically left-wing position, for example, a party can signal that it wants to defend the interests of the working class or the poor. There is ample evidence of such programmatic linkages for both economic and moral/religious issues (Evans and Tilley Reference Evans and Tilley2012; Jansen et al. Reference Jansen, De Graaf and Need2012).
A recent but rapidly growing literature draws attention to a third strategy that parties can use to reach out to groups: appealing directly to them in campaign communication (e.g., Dolinsky Reference Dolinsky2023; Huber Reference Huber2022; Stuckelberger and Tresch Reference Stuckelberger and Tresch2024; Thau Reference Thau2019, Reference Thau2021, Reference Thau2024). Such a strategy allows even politicians who are members of a different social group to ‘reduce social space’ between themselves and the members of the group they are targeting, with the ultimate goal of increasing support among this group (Alamillo and Collingwood Reference Alamillo and Collingwood2017). Group appeals often include references to policy positions and help to clarify to specific groups that the candidate or party shares their position on specific issues, but there is also a rhetorical element in signaling that one cares about a specific group (Berinsky et al. Reference Berinsky, de Benedictis-Kessner, Goldberg and Margolis2020). Thau (Reference Thau2018, Reference Thau2019) has documented that parties make extensive use of ‘symbolic group appeals’ in their party platforms. Such appeals, defined as ‘explicit statements that link a political party to some category of people’ (Thau Reference Thau2019, p. 65), are widely used in party manifestos, speeches, and press releases, with a growing diversity in terms of the groups appealed to (Dolinsky Reference Dolinsky2023; Huber Reference Huber2022; Stuckelberger and Tresch Reference Stuckelberger and Tresch2024; Thau Reference Thau2019). To clarify the concept, Table 1 provides selected examples of group appeals, taken from party manifestos of British parties. The examples illustrate some of the variation in the types of groups that parties appeal to.
Examples of group appeals in British party manifestos
In terms of the effects of group appeals on voting behavior, a limited number of studies provide indications that these strategies are electorally effective. Thau (Reference Thau2021, p. 18), shows that ‘working-class voters become more distinct from voters from other classes the more Labour emphasizes its working-class ties’ in their manifestos. This suggests that by appealing – even symbolically – to a specific group, parties strengthen their link with that group and gain electoral support among its members. Experimental work provides evidence for this role of symbolic group appeals in shaping citizens’ attitudes towards parties and their likelihood to support them. Drawing on vignette experiments that were embedded in surveys in Denmark and the United States, Robison et al. (2021) show that working-class respondents who saw a vignette that included a symbolic appeal to the working class were significantly more likely to support the candidate making the appeal than those in the control group.
From work that has studied the effectiveness of ethnic and gender-based campaigning, we know that efforts to appeal to groups are not always electorally rewarding, however (Kam et al. Reference Kam, Archer and Geer2017). More specifically, by catering to a specific group, candidates and parties might alienate the out-group (Hersh and Schaffner Reference Hersh and Schaffner2013; Ostfeld Reference Ostfeld2019). The effects of group appeals, however defined, are therefore likely to differ depending on whether the individuals who are exposed to a group appeal are members of the in-group or not.
The impact of group appeals on voting behavior – in the form of increased support among the group that is appealed to or a backlash among others – has been connected to social identity theory (Tajfel and Turner Reference Tajfel, Turner, Austin and Worchel1979). This theory conceives of people’s sense of who they are as shaped by their membership of specific social groups and the contrast with members of other groups. Group appeals tap into these distinctions between groups in society. By doing so, the appeals render citizens’ in- and out-group memberships more salient in political considerations. This, in turn, strengthens the impact of these memberships on vote choice (Alamillo and Collingwood Reference Alamillo and Collingwood2017; Kam et al. Reference Kam, Archer and Geer2017; Ostfeld Reference Ostfeld2019).
These in- and out-group dynamics, as described by social identity theory, inform our main hypotheses about the impact that group appeals have on electoral support:
Hypothesis 1: Group appeals lead to greater electoral support among the group that is appealed to (i.e., the in-group).
Hypothesis 2: Group appeals lead to lower electoral support among those who are not in the in-group.
Among whom are group appeals effective? The conditioning role of social identity strength and perceptions of deservingness
While important, the distinction between in- and out-group members is only the first step in accounting for the effects of group appeals. We also expect variation within these groups to condition such effects. Specifically, we theorize that differences in the strength of individuals’ social identities and in their perceptions of the deservingness of the group that is appealed to result in heterogeneity in voter responses to group appeals. Variations on these factors may, thus, account for the seeming contradiction between Hersh and Schaffner (Reference Hersh and Schaffner2013) and Robison et al. (2021), who, respectively, find and don’t find backlash effects among members of the out-group to a given group appeal.
First, and in continuation of the SIT-based arguments presented above, we argue that the effects of candidates’ group appeals – among the in-group as well as among others – will be conditioned by the strength of voters’ relevant social identities. Social psychology (Huddy Reference Huddy, Huddy, Sears and Levy2013; Tajfel Reference Tajfel1981) has long stressed the importance of subjective social identities for shaping in- and out-group dynamics. The basic intuition is that mere objective membership in a given group does not mean that this group membership is central to an individual’s self-concept (Huddy Reference Huddy, Huddy, Sears and Levy2013). Group identities and their strength are more informative in that regard, as they signal not only whether individuals consider themselves – subjectively – a member of the in-group, but also what significance they attach to that group membership (Tajfel Reference Tajfel1981). Work in the field of voting behavior that has built on the social identity literature provides ample evidence that the strength of identities is what drives the political effects of identities (Bartels Reference Bartels2000; Bornschier et al. Reference Bornschier, Häusermann, Zollinger and Colombo2021; Huddy Reference Huddy, Huddy, Sears and Levy2013; Huddy et al. Reference Huddy, Mason and Aarøe2015). Based on such insights, we expect that individuals who identify more strongly as a member of a specific group care more about this membership and will be more responsive to a candidate or party appealing to their in-group.
There are also reasons to expect the strength of social identities to moderate how individuals who are not members of the in-group respond to group appeals. Specifically, we argue that individuals strongly identified with out-groups will react more negatively to such appeals than those less strongly identified with out-groups. Previous work on the impact of social identities on voting behavior provides some underpinning for this intuition. For example, studies have shown that the effects of group threat are larger among individuals who identify more strongly with a group (Garand et al. Reference Garand, Qi and Magaña2022). This is also in line with social identity theory, which not only predicts that external threat from another group strengthens group solidarity but also that reactions to threat are more pronounced among those who have strong group identities (Huddy Reference Huddy, Huddy, Sears and Levy2013).
Based on these considerations, we formulate the following conditioning hypotheses:
Hypothesis 3a: Among in-group voters, stronger identification with their own group leads to a more positive group appeal effect.
Hypothesis 3b: Among out-group voters, stronger identification with their own group leads to a more negative group appeal effect.
Reactions to group appeals may also be grounded in other considerations than social identities, however. Thus, as argued by Conover (Reference Conover1988, p. 57), who draws on reference group logic, individuals’ thinking about social groups is dominated by ‘the desire to know who is getting what and whether they deserve it’. The key term here is deservingness, i.e., the perception of whether or not a group is getting its just desert. The importance of deservingness perceptions has been documented by studies of citizens’ welfare attitudes and policy preferences (e.g., Cavaillé and Trump Reference Cavaillé and Trump2015; Petersen Reference Petersen2012), and deservingness perceptions have also been shown to correlate with party preferences (Attewell Reference Attewell2021). There is also work that shows that perceptions of the deservingness of specific groups can influence group dynamics. Studying attitudes about income groups in the United States, for example, Piston (Reference Piston2018) shows that the belief that the poor have less than they deserve not only shapes attitudes about income redistribution but also voting (see also Attewell Reference Attewell2021; Conover Reference Conover1988).
Most scholars who examine perceptions of deservingness highlight the role of these perceptions for understanding individuals’ reactions to an out-group. However, if members of an in-group are convinced that they are unfairly treated, they should welcome attention from political candidates or parties even more. Consequently, we argue that perceptions of deservingness will moderate the effect of group appeals among both the in-group and others. Specifically, we hypothesize that perceiving a group as more deserving will move the group appeal effect in the positive direction. That is, if the baseline effect of the appeal is positive, its effect should increase in size. If the baseline effect of the appeal is negative, deservingness should bring the coefficient closer to zero and potentially turn it positive. We thus formulate the following hypothesis:
Hypothesis 4: The higher the perceived deservingness of the group that is targeted by a group appeal, the more positive the group appeal effect.
We note that we do not have a priori expectations about the relative importance of the moderation effects of identity strength and deservingness, respectively. One might speculate, however, that identity strength is more important among in-group members, while out-group members react more on the basis of their deservingness perceptions (this consideration was not pre-registered, though). Having laid out these expectations, we now turn to our two empirical studies, presenting the design and results from each of these in turn.
Study 1: Class appeals and other types of group appeals
Data and design
The main objective of Study 1 is to reproduce recent findings on class appeals and to start probing the effectiveness of group appeals for other major social categories. The design replicates and extends the approach of Robison et al. (2021), who examined the impact of appeals to the working class by means of vignette experiments fielded in online surveys in Denmark and the United States.Footnote 2
We embedded experimental vignettes in an online survey among a sample of 1,500 British respondents of 18 years and older.Footnote 3 The survey was administered by YouGov between 21 and 27 January 2021 and is representative of the British population in terms of gender, age, race, and education. By focusing on the United Kingdom, we extend the findings of Robison et al. (2021) to a third setting. Furthermore, in Britain the role of class identities is known to be very strong (Evans et al. Reference Evans, Stubager and Langsæther2022). Exploring whether other types of appeals have effects that are similar to those of appeals to class groups in Britain can therefore be considered a hard test.
Respondents in Study 1 were randomly assigned to one of five groups: a control group and four group appeal treatment groups. All respondents saw the following baseline information:
Please carefully read the following information about a hypothetical candidate in a UK parliamentary general election.
[David Smith/Susan Smith] is running for a seat in the UK House of Commons as a member of the [Labour/Conservative] Party.
The introduction of the hypothetical candidate – which was the only information in the control – randomizes the candidate’s gender (either David or Susan Smith) and party affiliation (Labour or Conservative Party). Including this information provided respondents in the control group with some information to allow them to form an opinion about the candidate.
For respondents in the treatment groups, the baseline vignette continued with a group appeal. The working-class vignette, for example, looks as follows:
Working class vignette
[David Smith/Susan Smith] recently said, ‘too much attention has been given to businesses in recent political debates. We in the [Labour/Conservative] Party believe it is time for politicians to prioritize workers. As a Member of Parliament, I will work to better represent the interests of workers’.
Similar vignettes were included to appeal to people living in rural areas or small towns, to non-university graduates, and to young people. The full wording of these alternative vignettes is shown in online Appendix A.Footnote 4
The treatments thus consist of a specific group that the candidate symbolically appeals to. By symbolic, we mean that the candidate does not take a policy position but simply mentions that they will work to better represent a specific group (in opposition to an out-group). These vignettes closely follow the wording of the class appeal treatments of Robison et al. (2021), the only difference being the extended list of specific in- and out-group pairs that the candidate appeals to across treatment conditions.
Immediately following the vignettes, respondents were asked to rate a candidate with political views like those of the hypothetical candidate presented in the vignette on a scale from 0 (= thinks very poorly of him/her) to 10 (= thinks very highly of him/her). The dependent variable again follows the design of Robison et al. (2021), corresponding to their Study 1.Footnote 5 This design allows examining the effect of different group appeals on candidate support separately – implying that we are not assessing the possibility that different group appeals have a cross-cutting impact. We return to this point in the discussion section of the article.
Estimation approach
We hypothesized that individuals’ reactions to a group appeal are a function of whether they are a member of the in-group (H1) or not (H2). In line with H1, we expect that for those who are in-group members, exposure to a group appeal will increase candidate support among the group that is appealed to, in comparison to the control group. Furthermore, H2 predicts that among respondents who are not members of the in-group, exposure to the treatment will decrease candidate support in comparison to the control group. To examine the effect of the group appeals among in- and out-groups, we categorize them as either a member of the in-group or not using their responses to a series of survey questions measured before the experiment.Footnote 6
To test H1 and H2, we regress respondents’ rating of the candidate on whether or not they are members of the in-group, whether they received an appeal, and the interaction of these two. Dummies for candidate party and gender were added to increase precision (see online Appendix B for the estimation equation). We estimate separate models for each of the four group appeal treatments.
Results
We focus on the average marginal effect of exposure to a group appeal (versus the control) among members of the in-group and others. Full estimates can be found in online Appendix C. We start with the working-class appeal, which is a close replication of Robison et al. (2021). In Figure 1, panel (a) shows the results for this appeal. The right-hand coefficient in the panel shows an effect that is consistent with H1. That is, among respondents who categorized themselves as working class, exposure to the candidate appealing to the working class significantly increases their evaluation of the candidate, relative to the control (i.e., a positive and significant AME). The estimated effect of an appeal to the working class among those who do not self-identify as working class (the left-hand coefficient in the same panel), in contrast, does not show the expected pattern. Rather than a negative effect (H2), we find a positive effect. Workers do tend to respond more positively than others, however.Footnote 7
Impact of exposure to a group appeal among in-groups and others: Study 1 results.
Note: Circles indicate the average marginal effect of exposure to a group appeal versus being in the control group. Spikes show 95% confidence intervals.
The effects of the group appeals are sizable. For the in-group, the estimated effect of 1.61 corresponds to a 0.70 standard deviation increase in support for the candidate, while support among those not in the in-group increases by 0.39 standard deviations. These results for the appeal to workers closely mirror the findings of Robison et al. (2021). They also failed to find evidence of a backlash effect of a working-class appeal among non-working-class voters. Overall, the results for the working-class treatment thus suggest that appealing to the working class is potentially an electorally rewarding strategy. But can the effectiveness of group appeals be extended beyond a strongly identity-based and politicized cleavage like the class cleavage?
Panel (b) in Figure 1 shows the results for the second vignette, where a candidate appeals to people living in rural areas and small towns. This type of appeal affects candidate support in ways that are similar to the working-class appeal. Among respondents who live in small towns or rural areas (i.e., the in-group), the rating of the candidate is significantly higher than in the control. The substantive size of this effect, furthermore, is very similar to that for the working-class appeal. The results for the out-group are also in line with the working-class appeal. Appealing to people living in rural areas or small towns does not produce a backlash in candidate support among urban residents but tends to be positive (though insignificant).
The effectiveness of appeals to educational groups is shown in panel (c) of Figure 1. Results suggest that the lower educated (those who finished education before 20) rate candidates significantly more positively when they appeal to non-university graduates compared to the control. Furthermore, and just as for working-class and rural appeals, we also find a positive (but insignificant) effect among the out-group on the education appeal.
Panel (d) concerns the appeal to young people (in opposition to the elderly). To examine how this appeal alters candidate evaluations in the in- and out-group, we contrast the reactions of voters under 35 and all others. The picture that emerges differs from that of the other group appeals. First, there are no indications that young respondents (as defined here) react positively to a candidate appealing symbolically to young people. The results, therefore, are not in line with H1. Second, for this specific type of appeal there is evidence of a backlash effect (H2). As suggested by the left-hand coefficient in panel (d), compared to the control condition, those who are over 35 years old react significantly negatively to the candidate appealing to the young.
Summary and discussion of Study 1
The results from Study 1 replicate previous findings that working-class appeals can shape candidate evaluations but also suggest that the effectiveness of group appeals might not be limited to class-based appeals. Symbolic appeals to people living in rural areas and the lower educated can also increase candidate support among those groups, and, as was the case for class, these appeals seem to come at limited costs in terms of how out-groups respond. The pattern is the exact opposite for an appeal to young people, which is negatively received by out-groups and does not boost support among the young.
We tested the robustness of the results of Study 1 in a number of ways. First, our main conclusions hold when examining the effect of group appeals with respect to vote likelihood instead of our primary measure of candidate support. As online Appendix E shows, effects are generally weaker when analyzing vote likelihood. However, the results still suggest positive in-group effects for appeals to the working class, people in rural areas, and non-university graduates, while the hypothesized backlash against such appeals among those who are not the in-group is absent (the tendency for these out-groups to respond positively disappears, however). Further, for candidates appealing to young people, the vote likelihood results reproduce both the negative out-group response and the insignificant in-group response found in the main analysis.
Second, while we did not have a priori expectations about whether group appeals would have different effects based on the party of the hypothetical candidate, online Appendix F shows the results broken down by the party of the candidate (i.e., Labour or Conservative). While the estimates are not identical for the two parties, we cannot discern a systematic pattern in terms of how respondents react to appeals from Labour versus Conservative candidates. It does appear, however, that out-groups react more positively to a Conservative candidate appealing to the working class or lower educated than to a Labour candidate doing so.Footnote 8
Third, we explored the age-based appeals further by using more fine-grained age categorizations. As online Appendix D shows, the lack of electoral support among the in-group (i.e., young people) is not driven by the somewhat arbitrary cut-off of respondents being 35 years or less, while the backlash effect is concentrated among the oldest age groups (particularly 65+).
Overall, Study 1 offers support for H1 and much less so for H2. This suggests that, while group appeals’ effectiveness does not seem limited to class, there is important variation in how voters react to different types of appeals. Potentially, the absence of in-group effects for the appeal to young people could reflect that age-based identities are generally weak. Further, the lack of negative out-group reactions to appeals to the working class, rural people, or the lower educated could be driven by the fact that these groups are generally perceived as deserving of better conditions in life. Study 2 examines these sources of heterogeneity in how voters respond to different types of group appeals.
Study 2: The scope of group appeals and heterogeneity in voter responses
Experimental design and data
The data for Study 2 come from an online survey among 3,200 British respondents over 18 years old.Footnote 9 The data collection was administered by YouGov, and the fieldwork ran from 13 to 20 June 2022. Before collecting the data for Study 2, we preregistered our design, expectations, and analysis plan on OSF.Footnote 10 The objectives of Study 2 were to reproduce the broad applicability of group appeals indicated by Study 1 and to test the two hypotheses about social identity strength and deservingness perceptions.
The basic setup of the experimental design for Study 2 differed from Study 1 in two ways. Most importantly, treated respondents were exposed to multiple appeals. Specifically, one-third of all respondents (N = 1,054) were assigned to a control arm. This group was told that they would see a hypothetical candidate (as in Study 1) and then received the following information (the candidate’s party was randomized):
Control group vignette
The candidate is a member of the [Labour/Conservative] Party.
The remaining two-thirds of respondents (N = 2,146) were assigned to a treatment arm and were told that they would see seven different candidates. The order in which the candidate vignettes were shown to respondents was randomized. These vignettes included appeals to the working class, to people living in rural areas and small towns, to non-university graduates, to women, to the elderly,Footnote 11 and to ethnic minorities. A seventh vignette consisted of a broad appeal to all groups, a type of appeal on which we elaborate in the concluding section of the paper. The vignettes are presented in online Appendix G. Here, we illustrate the vignettes with an appeal to people living in rural areas and small towns:
Rural area vignette
The [Xth] candidate is a member of the [Labour/Conservative] Party and recently said, ‘We in the [Labour/Conservative] Party believe it is time for politicians to prioritize people living in rural areas and small towns more. As a Member of Parliament, I will work to better represent people living in rural areas and small towns’.
Following each vignette (one vignette for the control arm, seven vignettes for the treatment arm), respondents were immediately asked, ‘How would you rate a candidate like this on a 0 to 10 scale? 0 means that you think “very poorly” of them, and 10 means that you think “very highly” of them’. These candidate ratings are the dependent variable in the analyses for Study 2.Footnote 12
Showing seven vignettes to respondents in the treatment arm has two advantages. Methodologically, it maximizes power because we reuse the same respondent multiple times (standard errors are clustered where appropriate). Substantively, it introduces a response situation where respondents are exposed to various group appeals, which is likely closer to how political communication unfolds in practice. We acknowledge, however, that this design also has limitations. For example, respondents are exposed to group appeals sequentially, rather than simultaneously, precluding an analysis of how group appeals might interact (we discuss this in the concluding section). Further, given that the vignettes are very similar, showing seven different vignettes to respondents in the treatment group might result in demand effects. We therefore examine how the order of the vignettes affects results in our robustness analyses.
The second way that Study 2 differs from Study 1 is that the vignettes are shorter and focus on the group that is appealed to – without contrasting them to a specific out-group. Our inclusion of a contrast in Study 1 followed the design of Robison et al. (2021). However, because it is possible that the initial effects are driven by both parts of the statement (i.e., the appeal to an in-group and a negative reference about an out-group), and to reduce the burden on respondents in the treatment arm, we pilot-tested whether a shorter and simplified version of the class vignette only appealing to an in-group (and omitting candidate gender) had similar effects as the original vignettes and found it does (see online Appendix H). We thus went for the simpler setup for all the group categories in Study 2.
Using the data from Study 2, we pursue tests of all hypotheses, both the main ones and the hypotheses that concern conditional effects. The main independent variables for testing Hypotheses 1 and 2 are a variable identifying whether a respondent is randomly assigned to a treatment that includes an appeal targeted at a specific group and variables that identify whether the respondent is a member of the targeted group (i.e., in-group) or not. To identify whether respondents’ group membership implies they are targeted by the appeals, we rely on standard measures of each of the socio-demographic characteristics involved in the appeals. For details, see online Appendix I.Footnote 13
To test Hypotheses 3 and 4, we require measures that capture how strongly respondents identify with the in-group or a group that is not targeted by the appeal. To capture identity strength, we asked respondents to indicate how important their class, ethnic or racial background, place of living, gender, age group, and educational background were to their sense of who they are. Respondents were asked to indicate the importance of each of these characteristicsFootnote 14 on a scale from 0 to 10, where 0 means ‘not important at all’ and 10 means ‘very important’.
Furthermore, testing Hypothesis 5 requires indicators of how deserving respondents think different social groups are. We asked respondents to indicate, for ten different groups (including those appealed to), whether they thought that most people in each of the groups have poorer life conditions than they deserve, better life conditions than they deserve, or just about the life conditions that they deserve? Responses were given on a 0–10 scale, and answers were reverse coded such that higher values correspond to a more deserving group (i.e., the group has poorer life conditions than they deserve).Footnote 15 Online Appendix I provides details on this question, including the distribution of response patterns for all items.
We measured respondents’ socio-demographic characteristics, their identity strength, and their deservingness perceptions before the experiment. Just after the measurement of the socio-demographic variables and the key moderators, the questionnaire included a political knowledge quiz that served as a distractor task and that should reduce the effects of priming social identities on the treatment effects (see also, Anson Reference Anson2018; Kam Reference Kam2007).
Estimation approach
To test H1 and H2, we estimate linear models similar to those in Study 1 (see online Appendix B).Footnote 16 We first assess effects for each specific group appeal vignette, each time contrasting responses in the treatment arm with those in the control. In a second step, we estimate a pooled model to obtain an overall estimate of the effect of the group appeal among the in-group and others, respectively.
To test Hypotheses 3, 4, and 5, we have to test whether the effects of the group appeals as well as their interaction with respondents’ group memberships are conditioned by the strength of respondents’ social identities and their perceptions of deservingness. Empirically, we do so by adding to the equation three-way interactions and all constitutive terms (see online Appendix B). For testing Hypotheses 3 and 4, we thus add a three-way interaction between identity strength and the appeal treatment and group membership variables. For testing Hypothesis 5, we add a three-way interaction between deservingness perceptions and the appeal treatment and group membership variables.
To facilitate the interpretation of the effects, we calculate the average marginal effect (AME) of the group appeal treatment as the strength of group identity or deservingness increases. We do so for in-group members and those who are not in the in-group separately. According to Hypothesis 3, the AME of the group appeal treatment will increase as identity strength increases for in-group members. Hypothesis 4 supposes that the AME of the group appeal treatment will decrease as identity strength increases for those not in the in-group. According to Hypothesis 5, the AME of the group appeal treatment will become more positive as the group is considered more deserving. The results for the moderation effects presented in the paper are based on a pooled analysis of a stacked data matrix in which respondents are included as many times as there are group appeal vignettes. For all pooled analyses, we cluster the standard errors at the level of individual respondents.
Results
Figure 2 shows the effects of the group appeals for each of the six groups considered in Study 2 (detailed estimates can be consulted in online Appendix J). The setup follows that of Figure 1. The top row shows the effects for group appeals that were also considered in Study 1, i.e., appeals to the working class, to people living in rural areas and small towns, or to people without a university degree. We see that respondents reacted in largely the same way to these appeals as respondents in Study 1 did. For each of the appeals, the in-group responded positively to the appeal, rating the candidate making the appeal between 0.8 (for the education-based appeal) and 1.7 points (for the class-based appeal) more positively than the control group. Also in line with Study 1, for none of these appeals is there evidence of a backlash effect among respondents who are not targeted by the appeal – if anything to the contrary.
Impact of exposure to a group appeal among in-groups and others: Study 2 results.
Note: Circles indicate the average marginal effect of exposure to a group appeal versus being in the control group. Spikes show 95% confidence intervals. Detailed estimates are reported in online Appendix J.
The bottom row in Figure 2 shows the estimates for the three appeals that were not included in Study 1: appeals to the elderly, to women, and to ethnic minorities. We see that respondents 65 years old and older, women, and ethnic minorities all respond positively to a candidate appealing to them. Furthermore, the average marginal effects are sizable, ranging between 1.1 and 1.8. Thus, consistent with H1, the in-group effects show how candidates can use group appeals to increase their support among a specific social group.
Among those who are not members of the in-group, the effects shown in the bottom row of Figure 2 corroborate the initial evidence against H2. For none of these three types of group appeals is there a backlash among out-group respondents. On the contrary, for each of the three appeals, out-group members evaluate the appealing candidate more positively than in the control condition.
Having reproduced the result that each of the vignettes has the intended effect among in-group members, while also increasing candidate support outside the in-group, we now turn to a pooled analysis of the stacked dataset. In this dataset, each respondent is included six times, for each of the six comparisons between a group appeal condition and the ‘no appeal’ control (N=19,036). In a first step, we estimated the overall in-group and out-group effects (i.e., pooling the information in Figure 2). This model replicates the positive and significant effect of group appeals both among in-group (the estimated effect is 1.2 points) and out-group (the estimate is 0.6 points) members (see online Appendix K). Again, these effects provide strong support for H1 while failing to support H2. To test H3, H4, and H5, we expand the baseline pooled model to include three-way interactions with the measures of identity strength and deservingness perceptions, respectively.
Figure 3 shows the average marginal effect of assignment to a group appeal versus the control on candidate ratings as a function of the strength of a respondent’s identity.Footnote 17 Panel (a) shows the relevant effects for members of the out-group, and panel (b) concerns in-group members. We first look at in-group members, for which we hypothesized that a stronger identification would lead to a more positive group appeal effect (H3). As the upward-sloping line in panel (b) suggests, the effect of being assigned to an appeal to one’s in-group increases somewhat as the group gains importance for respondents’ sense of who they are. Specifically, if a group membership is ‘not important at all’, being assigned to a group appeal improves candidate ratings by about 1.0 point compared to the control. If the group membership is ‘very important’, the electoral effect is about 1.4. This conditional effect of identity strength, however, is estimated quite imprecisely and is not statistically significant. Vignette-specific estimates reported in online Appendix N furthermore show that this conditional effect is mostly driven by the appeals to people living in rural areas and small towns, women, the elderly, and ethnic minorities.
Impact of exposure to a group appeal among in-groups and others, conditional on identity strength.
Note: The solid line indicates the average marginal effect of assignment to a group appeal versus the control. The shaded area indicates 95% confidence intervals. Histogram indicates the distribution of identity strength. Detailed estimates are reported in online Appendix L.
Panel (a) in Figure 3 focuses on effects among out-group members. For them, we hypothesized that stronger identification with their own group would lead to a more negative group appeal effect (H4).Footnote 18 The results offer mixed support for this hypothesis. On the one hand, the (positive) effect of being assigned to an appeal to a group that one is not a member of is significantly reduced the stronger one’s out-group identity is. Specifically, for a non-in-group member for whom their out-group membership is ‘not important at all’, being assigned to a group appeal is associated with a 1.0 point higher candidate rating. This effect is reduced to 0.1 and becomes indistinguishable from zero when the out-group-membership is ‘very important’ to one’s sense of who one is. Identity strength thus significantly moderates the effect of a group appeal among those who are not members of the in-group. On the other hand, even among those whose out-group membership is ‘very important’, a group appeal does not lead to a backlash against the candidate making the appeal. When analyzing each of the vignettes separately (see online Appendix N), it becomes clear that this holds quite generally. The only vignette for which there are indications of a significantly negative out-group effect under strong identity strength is the appeal to ethnic minorities.
Finally, we turn to the moderating effect of deservingness perceptions. We test the hypothesis that the higher the perceived deservingness of the group that is targeted by a group appeal, the more positive the group appeal effect should be (H5). We proposed one general hypothesis for both in- and out-group members, but we allow for the possibility of different effects by assessing the moderating effect of deservingness perceptions on the impact of group appeals for both groups separately. Figure 4 shows the results for out-group members in panel (a) and in-group effects in panel (b).Footnote 19
Impact of exposure to a group appeal among in-groups and others, conditional on perceptions of deservingness.
Note: The solid line indicates the average marginal effect of assignment to a group appeal versus the control. The shaded area indicates 95% confidence intervals. The histogram indicates the distribution of deservingness perceptions. Detailed estimates are reported in online Appendix M.
Both panels plot the average marginal effect of being exposed to a group appeal vis-à-vis the control on candidate ratings as a function of how deserving the group that is appealed to is perceived. The deservingness measure is coded such that higher values signify that the group is perceived as deserving better conditions in life, while low values signify the group is perceived as one that already has better life conditions than they deserve.
Figure 4 shows that deservingness perceptions strongly moderate the effect of group appeals – among in- and out-group members alike. Starting with the in-group (panel (b)), the reward for a candidate appealing to the group is 2.4 points when deservingness perceptions are at the maximum value. This effect is about double the size of the overall effect of in-group appeals when not accounting for heterogeneity in the effects. At the other end of the range of deservingness values, in-group members who believe that their group already has better life conditions than they deserve punish a candidate who appeals to the group. At a deservingness value of 0, this effect is −0.7 and statistically different from zero. It should be kept in mind that there are very few respondents who rate their own group at zero on the deservingness measure.Footnote 20
As panel (a) in Figure 4 shows, the moderating effect of deservingness perceptions works in a similar way, although somewhat more strongly among out-group respondents.Footnote 21 The effect of being exposed to an appeal to a group that one does not belong to ranges from −1.8 when the appeal concerns a group that scores 0 on the deservingness scale to 2.0 when the group receives the maximum deservingness score. Thus, depending on how one perceives the group that is appealed to, a candidate making an appeal can either lose or win up to 2 points in popularity among voters who are not themselves members of the in-group.
The effects in Figure 4 are not only substantively very large, but they also clarify the conditions under which candidates can expect a backlash from relying on group appeals: when candidates appeal to groups that are perceived as low-deserving, out-group members punish the appeal. Specifically, if deservingness values are 4 or below, the effect of exposure to a group appeal is significantly negative among out-group members.
The vignette-specific analyses of the moderating effect of deservingness perceptions in online Appendix O show that there is variation in the strength of this moderation effect, but that the finding that deservingness perceptions are a strong moderator of reactions to group appeals applies for all appeals.
Summary and discussion of Study 2
Study 2 finds an overall positive effect of group appeals among both in- and out-group members across six politically salient social groups and thus supports H1 but not H2. Crucially, however, Study 2 sheds light on the sources of variation in how citizens respond to candidates making a group appeal. Social identity strength matters somewhat (H3), especially among out-group members (H4). Notably, the positive out-group effects of group appeals seem concentrated among out-group members for whom their out-group membership is not an important part of their sense of who they are. The results furthermore show that deservingness perceptions strongly moderate reactions to group appeals (H5). This holds for in- and out-group members alike, although the conditioning effect of deservingness perceptions is somewhat more pronounced among out-group members. The results also clarify that backlash effects do occur, but they are concentrated among out-group members who think the group appealed to already has better life conditions than it deserves.
We tested the robustness of the results of Study 2 in a number of ways. First, to be able to more clearly probe the unique effects of identity strength and deservingness perceptions as compared to simple likes or dislikes of groups (as per reference group theory, see, e.g., Conover Reference Conover1988; Huber et al. 2025; Piston Reference Piston2018), our survey also measured group affect using standard like/dislike questions (on 0–10 scales). In online Appendix Q, we show that group affect also moderates respondents’ reactions to group appeals. Most importantly, however, both the identity strength and deservingness moderation effects hold when we account for heterogeneity based on respondents’ affect towards the groups that are appealed to.
Second, we examined whether there is heterogeneity in the effects depending on the party of the candidate that is shown in the vignette (as for Study 1). These analyses show that the effects are mostly consistent across parties, though effects are generally stronger when the appeal comes from a Conservative candidate (see online Appendix R).
Third, respondents who were assigned to the treatment arm were shown 7 vignettes in which appeals and the candidate’s party varied but otherwise were very similar in terms of the wording. We explored whether effects perhaps weaken or strengthen after exposure to multiple vignettes and found that while the group appeal effect appears somewhat larger for whichever vignette comes first, there are no statistically significant order effects (see online Appendix S).
Finally, as our design is tailored to assess the additional effect of a group appeal on candidate evaluations (compared to when respondents only have basic candidate information), we do not explicitly contrast the effectiveness of group appeals with other types of appeals. To contextualize the results, however, online Appendix T leverages the inclusion of the broad appeal vignette in Study 2 to compare the group appeal effects with the effects of candidates appealing broadly. The results indicate that broad appeals substantially increase candidate support, but unlike group appeals, they do so independently of respondents’ in- and out-group status. A second way to contextualize the importance of group appeals is by contrasting the impact of being assigned to an in-group appeal with assignment to an in-party candidate. In a non-preregistered, and therefore exploratory, analysis reported in online Appendix U, we show that the effect of being assigned to an in-group appeal (versus an out-group appeal) is about half the size of being assigned to a candidate from the in-party (versus a candidate from the out-party).
Conclusion
When choosing candidates or parties, voters are often guided by their social characteristics, group memberships, and identities. Group voting, however, does not emerge in a vacuum but can be mobilized by politicians. One tool that parties have to clarify their stance on group conflicts is to appeal explicitly to specific groups. Research that analyzes party communication has found that parties rely strongly on such strategies, e.g., by including large numbers of appeals to groups in their manifestos (Dolinsky Reference Dolinsky2023; Huber Reference Huber2022; Riethmüller and Franzmann 2025; Huber and Haselmayer Reference Huber and Haselmayer2025; Stuckelberger and Tresch Reference Stuckelberger and Tresch2024; Thau Reference Thau2019).
We know less about the effects on citizens of such appeals and the mechanisms that explain whether citizens reward or punish candidates or parties when they appeal to groups. This paper contributes to our knowledge about the individual-level effects of group appeals, with a focus on two important gaps in the literature. First, existing research on group-centric campaigning is fragmented, either focusing on isolated group categories or using differing concepts like ‘social bridging’ (Alamillo and Collingwood Reference Alamillo and Collingwood2017) or ‘targeted ads’ (Hersh and Schaffner Reference Hersh and Schaffner2013), often bundling together a range of elements that might appeal to groups (e.g., policy, symbolic, descriptive representation). Further, work focusing more narrowly on the symbolic group appeals highlighted in recent party research has so far been limited to social class and, more recently, place-based appeals (Haffert et al. Reference Haffert, Palmtag and Schraff2024). This has left open the question of whether symbolic appeals to other groups similarly shape candidate evaluations and vote choices. Second, we know little about the conditions under which in-group and out-group voters respond positively, or negatively, to candidates appealing to a specific group. Some studies find a backlash from out-groups (Hersh and Schaffner Reference Hersh and Schaffner2013), while others do not (Robison et al. 2021), and we lack a general explanation as to why such reactions vary across groups.
Our study has used vignette experiments embedded in surveys among British respondents to address these issues and makes three contributions to the literature. First, our results indicate that candidates can successfully appeal to a wide and varied set of social groups. With only one exception (appealing to young people), we find that in-groups reward a candidate when they appeal to them. Second, although in-group and out-group reactions clearly differ, we find that appealing to a specific group hardly produces negative effects among those who are not members of the group. For most of the appeals that we studied experimentally, out-group members reward a candidate who is making an appeal to another group – though not to the same extent as in-group members. The lack of strong backlash effects in our experiments, while noteworthy, is likely a consequence of the types of groups that we focused on. These are generally seen as deserving groups, and we show that deservingness perceptions are an important moderator of respondents’ reactions to group appeals. This brings us to our third and final contribution: beyond the average effects, we find that there is a substantial amount of heterogeneity in responses to group appeals. Social identity strength, and perceptions of deservingness in particular, condition how citizens respond to a group appeal. And under certain conditions, a candidate is punished when they appeal to a group. This holds especially when a group is perceived as low-deserving, where we find that group appeals can lower candidate ratings by up to 2 scale points – a substantial drop in popularity.
Our study is not without limitations. First, our data come from a single case: the United Kingdom. Given our focus on individual-level mechanisms, however, we have no reason to assume that the patterns we observed would be fundamentally different elsewhere. If anything, the UK – a setting where class politics has deep roots (Evans and Tilley Reference Evans and Tilley2012) – is likely a hard case to find effects of similar size for other types of appeals. Nonetheless, our results show that a wide variety of group appeals shape respondents’ evaluations of candidates. The fact that the effects of working-class appeals in our study are very similar to what previous work on such appeals in Denmark and the U.S. has shown also strengthens our confidence that the findings can be generalized outside the British context. Second, by focusing on sources of heterogeneity based on citizens’ identities and attitudes, we have left open the possibility that features of the candidate or the party making the appeal also condition effects. Additional analyses in online Appendix F suggest some amount of variation in this regard. Third, we have studied the effect of specific group appeals in isolation by contrasting one specific appeal at a time to the control. In real life, however, candidates might well appeal to different groups simultaneously, and different candidates and parties can compete for a group’s attention and loyalty by simultaneously appealing to the same group. Under such conditions, the effects of a group appeal are likely more muted than what is suggested by the effect sizes reported in this article. Finally, our results on group appeals’ effects raise the question of how group appeals compare to, and interact with, other kinds of electoral strategies. While the relative effectiveness of group and policy appeals has received some attention (Finseraas et al. Reference Finseraas, Heath, Langsæther and Smeets2025; Robison et al. 2021), there are many other ways that political elites appeal to voters. Thus, as these points suggest, more work is needed on the role and the effects of group appeals, and we encourage the field to go further in that direction.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1475676525100170.
Data availability statement
All data and code needed for reproducing and replicating the analyses reported in this paper are available on Harvard Dataverse at https://doi.org/10.7910/DVN/AZEKFZ.
Acknowledgements
Previous versions of this paper were presented at the 2021 Annual Meeting of the American Political Science Association, at the Electoral Chairs Seminar at the Université de Montréal, at the Government Department Speaker Series of the University of Essex, the Political Behavior Colloquium of the European University Institute, at the Center for Comparative and International Studies (CIS) Colloquium in Zürich, at the LSE Political Behavior seminar series, at the 12th Annual Conference of the European Political Science Association, Prague, at the 2022 ECPR General Conference, Innsbruck, at the ‘Changing Voters, Changing Parties: Re-Defining the Electoral Landscape in Europe’ workshop, 20–21 October 2022, University of Pittsburgh, at the 2022 Toronto Political Behavior Workshop, at the UNC Comparative Working Group speaker series and at the GEPOP seminar of the University of Gothenburg. We thank participants and discussants on all these occasions for their feedback and helpful suggestions.
Funding statement
This project was funded through the Canada Research Chair program (grant number 950 - 233132).
Ethical standards
Both surveys used in this paper received ethics clearance from the Université de Montréal. Respondents were asked to provide informed consent for their participation. In the consent form, they were informed about the broad objectives of the survey, that anonymity would be preserved, and how the data would be stored. Respondents were also given contact details of the ombudsman of the Université de Montréal and the researchers.
Open access
