Similarity plays a crucial role in impression formation: People are drawn to others whose values and attitudes are similar to their own (Byrne, 1971; Henry & Reyna, 2007), and dissimilar attitudes can be a rather strong interpersonal repellant (Brandt et al., 2014; Byrne, 1969). As political ideology is a prominent pointer to value dissimilarity (Motyl, 2014; Motyl et al., 2014), it is unsurprising that it influences interpersonal treatment. For example, Mallinas et al. (2018) found that for potential romantic partners, initial impressions changed after people discovered the target’s political background. Specifically, discovering a politically similar orientation sparked increased interest, while dissimilarity caused reduced interest. Moreover, in the U.S., marriage across party lines is rare (Rosenfeld et al., 2011), and close friendships exhibit a similar pattern (Pew Research Center, 2017).
Because attitudes and values are closely tied to a person’s identity and sense of personal worth (Hitlin, 2003), people might feel uncomfortable or offended when those values are challenged (Henry & Reyna, 2007). By associating themselves with others who support their beliefs and values, people receive validation, and become more likely to retain them over time (Brandt et al., 2014). Moreover, similarity could signal trustworthiness (Koch et al., 2020) and kinship (Park & Schaller, 2005), and to many individuals, group memberships such as political affiliation can be an important social identity (Simon & Klandermans, 2001), particularly in democratic societies where people are highly educated and engaged in politics (Caprara & Vecchione, 2017). Approaching environments that contain ideologically similar others can therefore be a means of satisfying the need to belong (Motyl et al., 2014). Discriminating dissimilar people could serve not only to protect one’s own values, but also to demonstrate to the ingroup that one rejects the values of the outgroup. Because ideologies stem from deeply held beliefs, they can elicit strong emotions, which provide justification for prosocial or antisocial behaviors (Johnson & Roberto, 2019). An additional contributor might be the perceived controllability of political affiliation: People choose to be liberal or conservative, making it easier to prescribe blame or credit to their group membership, compared to categories that are biologically determined (Iyengar & Westwood, 2015). Indeed, Brandt (2017) found that perceived ideology was a particularly strong predictor of prejudice. There is also evidence suggesting that preferences for the like-minded has escalated into increased prejudice against political opponents in several western countries (Palonen, 2009). In the US, dislike for the opposing party has been rising continually, while the level of liking of one’s own party has remained stable (Iyengar et al., 2012). Democrats and Republicans also overestimate each other’s ideological extremity, which may fuel mutual animosity (Parker et al., 2021; Westfall et al., 2015).
The Organizational Context
In workplace settings, there can be several consequences of ideologically driven self-selection and differential treatment. People within an organization become more homogenous over time (De Cooman et al., 2009), and employees in an organization that financially supports an opposing political party are more likely to turn over (Bermiss & McDonald, 2018). Moreover, employees can become excluded or unfairly treated if they do not share the organization’s values (Stone-Romero et al., 2003). Indeed, analogous to individuals’ ideology, organizations can be plotted on a liberal-conservative dimension, and ideologically skewed organizations are prone to make ideologically consistent decisions (Gupta et al., 2017).
Discrimination against job applicants and workers is a problem that entails negative consequences for individuals, organizations, and society more broadly. Hence, extensive research efforts have been devoted to discrimination based on demographic categories such as ethnicity (e.g., Quillian et al., 2019) and gender (e.g., Isaac et al., 2009). Organizational scientists have further studied whether demographic similarity between rater and target matters for work-related outcomes (e.g., Goldberg et al., 2014). However, research on discrimination based on beliefs or values is scarce. This study examines whether people discriminate against job applicants whose political orientation is dissimilar from their own, and whether they negatively stereotype them on two fundamental dimensions of impression formation – warmth and competence.
Hiring managers form impressions of applicants’ traits and skills based on their resumes, and these impressions guide judgments of employability (Burns et al., 2014; Sinclair & Agerström, 2020). When limited information is available, trait-based attributions based on social categories are especially likely, and ingroup – outgroup categorization generally leads to stereotyping (Stangor et al., 1992). Indeed, Koch et al. (2016) showed that people spontaneously stereotype groups based on political ideology, and that conservative-progressive beliefs were among the most important stereotype components, suggesting that political information is a basic aspect of interpersonal perception in contemporary post-industrial democracies. Furthermore, political ideology might become the basis for evaluating whether an applicant would fit into the organization’s culture (Johnson & Roberto, 2019).
Only a few studies have examined effects of applicants’ political affiliation in a hiring context. Inbar and Lammers (2012) found that about a third of their (liberal) social and personality psychologist sample admitted to being willing to discriminate against politically dissimilar applicants. These findings were subsequently extended to conservative as well as liberal academics from various disciplines (Honeycutt & Freberg, 2017). Another study (Iyengar & Westwood, 2015) found that both Democrats and Republicans showed tendencies to recommend someone from their own party to receive a scholarship, even in the case of a more qualified outgroup candidate. Moreover, Roth et al. (2020) found that perceived similarity to applicants who signaled their political affiliation in their Facebook profile was related to liking, which was in turn related to expectations about task performance and organizational citizenship behavior. Finally, Gift and Gift (2015) conducted a randomized field experiment with the correspondence testing method and found that real employers on the U.S. labor market offered fewer job interview invitations to applicants who signaled a minority political affiliation when applying for entry-level jobs: Republican applicants were more likely to receive callbacks in a conservative county than Democratic applicants, while the reverse was found for a liberal county. Although these findings are intriguing, the authors used data based on county voting patterns as a proxy for decision maker political affiliation. In addition to measuring political affiliation at the aggregate rather than the individual level, the fact that the aggregate analysis only concerned two counties makes it more difficult to rule out that other factors than the hiring manager’s political orientation account for the differential callbacks for Republican versus Democratic applicants.
We extend previous research by examining effects of rater-target political dissimilarity vs. similarity, focusing on hiring-related judgments but also two novel outcomes highly relevant for the workplace context: willingness to cooperate and socialize at work, and stereotyping based on warmth and competence. Warmth relates to perceived intent of a person/group and includes traits like helpfulness, friendliness, and sincerity, while competence relates to perceived capability and associated attributes such as ambitiousness and intelligence (Fiske et al., 2007). Stereotyping has been suggested as a potentially important mechanism for discriminatory behaviors in the workplace (Johnson & Roberto, 2019), and studies of whether out-partisan applicants are disliked in general or actually perceived as less competent have been called for (Gift & Gift, 2015). Stereotype content could reveal whether political opponents are perceived as less competent, less warm, or both, and whether they are discriminated against without the involvement of stereotyping. The stereotype content model (SCM; Fiske et al., 2002) proposes that a disliked outgroup that competes with the ingroup for resources should be stereotyped as low on both competence and warmth relative to the ingroup, and this is the pattern that we expect to find. Iyengar et al. (2012) found that targets were rated higher on perceived intelligence when raters’ political affiliation was similar to that of the target, while rater-target dissimilarity produced increased selfishness ratings. Our study adds to this literature by examining stereotype content in the context of judgments of politically dissimilar vs. similar job applicants.
Apart from hiring discrimination and stereotyping, negative treatment of out‐group members can take more subtle forms, not the least in organizational contexts. If political similarities attract and dissimilarities repel for friendships and romantic partners, this may also be true for relationships among coworkers. Subtle forms of negative interpersonal treatment could involve excluding someone socially or avoiding contact. For example, Koch et al. (2020) found that similarity in conservative- progressive beliefs had a stronger effect on cooperation than similarity in agency. We therefore hypothesize that employees will be less willing to socialize and cooperate with a potential co-worker when they learn that this person is politically dissimilar rather than similar.
The “traditional hypothesis” predicts that people holding conservative worldviews are more prone to protect their worldviews through prejudice compared to people with more liberal worldviews (see e.g., Ganzach & Schul, 2021). In contrast, the “worldview conflict hypothesis” expects people with both traditional and liberal worldviews to protect their worldviews by means of prejudice toward those with competing political views (Brandt & Crawford, 2020). Several recent studies suggest that negative intergroup attitudes are expressed on both sides of the political spectrum (e.g., Brandt et al., 2014; Mallinas et al., 2018). Thus, we expected both Republicans and Democrats to stereotype and discriminate against politically dissimilar applicants.
Study Design and Preregistration
We designed an experiment to test the hypothesis that negative treatment and stereotyping of applicants who express liberal vs. conservative political orientation will interact with the rater’s political affiliation (Republican or Democrat), such that people prefer applicants who express a political orientation similar to their own. Applicant political orientation was manipulated in a between-subjects design. We also varied whether the occupation was office assistant or IT-technician to be able to generalize our findings across two common occupations that vary in skill-level and are perceived to require different attributes related to warmth and competence (Strinić et al., 2022). Additionally, they are both realistic for our scenario where one’s employer is recruiting a new employee and asking for assistance in the screening process.
For reasons of transparency, we report how we determined the sample size, all data exclusions, all manipulations, and all measures in the study. The preregistration can be found in the Supplementary Materials.
Participants were recruited on Prolific and received compensation at an hourly rate of 7.05 British pounds. To achieve 90% statistical power to detect small effects of d = 0.4 for the t-tests (two-tailed, alpha level = .05) 266 Democrats and 266 Republicans for each of the occupations would be needed (i.e., 1064 participants in total). Because we estimated that up to 20% could fail the manipulation check, we added 213 to this number, planning to recruit 1,277 participants.
The initial sample consisted of 1,280 participants with equal representation of Democrats and Republicans according to information that they had previously provided when registering on Prolific (the participants were not aware of this pre-screening filter). We also required participants to be currently employed, 18-70 years old, fluent in English, and U.S. residents.
Because 264 participants (20.6%) failed the manipulation check (whether the applicant had mentioned left/liberal, right/conservative, or not mentioned political orientation), we excluded them according to the pre-registration. We based our groups on their current partisan self-categorization rather than the Prolific prescreening filter, due to the possibility that some participants might have changed their political affiliation since they signed up at Prolific. Although party identification tends to be highly stable over time within the context of the US biparty system, shifts sometimes occur (Jost et al., 2008). In our sample, 43 participants (4.2%) identified as neither Democrat nor Republican; these were therefore excluded. This procedure follows recommendations by Roth et al. (2020), who called for future studies to measure the actual political affiliation directly to avoid problems with proxies and imputation.
The final sample consisted of 973 participants, of which 511 identified as Democrats and 462 as Republicans. The two groups differed somewhat in terms of age (Democrats: M = 29.43, SD = 9.64, Median = 27; Republicans: M = 37.57, SD = 12.64, Median = 35) and therefore in work experience (Democrats: M years = 10.40, SD = 9.26; Republicans: M = 17.91, SD = 11.70), but had similar education level, with 59% having a university degree. The groups also had different gender compositions (Democrats: 17.4% men, 81.2% women, 1.4% other; Republicans: 42.9% men and 57.1% women). The participants’ self-placement on a scale where 1 means very liberal and 9 means very conservative confirmed that the Democrats were liberal (M = 2.26, SD = 1.25) and the Republicans were conservative (M = 6.95, SD = 1.26).
Materials and Procedure
The anonymous online-survey was created with Qualtrics software. After providing informed consent, the participants were informed that they would evaluate a job application, and received the following instructions:
“Please imagine that your current employer is looking to hire a full-time office assistant with administrative responsibilities. This person will be responsible for providing administrative support to ensure efficient operation of office, answering phone calls, scheduling meetings and supporting visitors, and carrying out administrative duties such as filing, typing, copying, scanning etc. He or she will also support teams by performing tasks related to organization and strong communication.”
“Please imagine that your current employer is looking to hire a full-time IT technician. This person will be responsible for diagnosing, repairing, and maintaining hardware and software components to ensure the smooth running of computer systems. His or her responsibilities will also include installing and configuring computer hardware and being the primary point of contact for IT support within the organization.”
All participants then read the following:
“Your employer asks you to help with the task of screening incoming applications. Considering that you will have to interact and collaborate with the person who ends up being hired for the job, it is in your best interest to find a suitable candidate. Please attend carefully to all of the information in the application.”
Next, they were randomly assigned to receive an application from either a liberal or a conservative applicant. The applications consisted of a personal letter and CV (available in the OSF repository, see Supplementary Materials). These were constructed based on applications that had previously been used in large field experiments (Carlsson & Eriksson, 2019). The applicant was a 31-year-old man named Thomas Miller whose educational background and previous work experience were based in Florida. The applicant’s political orientation was manipulated at the end of the personal letter:
“… I would describe myself as a social person who is used to a high work pace and who pays attention to detail. In my spare time, I like to be physically active and to socialize with friends. I am also passionate about politics and ideologically, I am conservative and sympathize with the political right [liberal and sympathize with the political left]. I hope to meet you for an interview. Thank you for your time and consideration.”
To encourage participants to pay attention to the manipulation, they had to remain with the application for at least 20 seconds before they could proceed. The participants then completed four outcome measures.
Hireability judgments were measured with Derous et al. (2009) job suitability scale: “Given all information you read about this applicant, how suitable do you believe this applicant is for this job?” (1 = Not suitable at all – 7 = Very suitable), “Given all information you read about this applicant, how likely is it that you would invite this applicant for a job interview?” (1 = Not likely at all – 7 = Very likely), and “Given all information you read about this applicant, what is your general impression of this applicant for the job?” (1 = Very negative – 7 = Very positive). These items were averaged and the internal reliability for this scale was Cronbach’s α = .90 for the assistant and .91 for the IT-technician).
Willingness to cooperate and socialize with the target was measured with three items: “If this applicant is hired for the job, to what extent would you like to cooperate with him?”, “If this applicant is hired for the job, to what extent would you like to get to know him better?”, and “If this applicant is hired for the job, to what extent would you like to talk to him during coffee breaks or lunch?” (1 = Not at all – 7 = Very much; Cronbach’s α = .90 for the averaged items, both occupations).
Stereotypes were assessed in terms of averaged ratings of how competent, talented, skillful, and ambitious (competence: Cronbach’s α = .92 for the assistant and .91 for IT-technician), and friendly, warm, well intentioned, and considerate (warmth: Cronbach’s α = .95 for both occupations) the applicant was perceived as (1 = Not at all – 7 = Extremely; previously used by Strinić et al., 2022). This was followed by demographic questions. The participants thereafter responded to the liberal – conservative self-placement question and to which of three alternatives (Republican, Democrat, or neither) best described their political affiliation.
Next, participants indicated whether the job applicant had mentioned political left/liberal orientation, political right/conservative orientation, or not mentioned it; this served as our manipulation check.
Finally, we measured willingness to cooperate and socialize with politically similar/dissimilar co-workers once more, this time by asking the participants to make an explicit comparison of their attitudes toward a conservative and a liberal co-worker. We measured these explicit attitudes by asking: “Generally speaking, if you find out that a new co-worker is politically liberal/left-wing, to what extent would you like to”: “cooperate with this person”; “get to know the person better”; “talk to the person during coffee breaks or lunch?” (1 = Not at all, 7 = Very much). The same question was asked regarding a conservative/right-wing co-worker; the order (liberal vs. conservative co-worker) was counter balanced and Cronbach’s α = .91 for both the liberal and conservative co-worker.
The dataset is available on OSF (see Supplementary Materials). All analyses followed the pre-registered plan, unless stated otherwise. The alpha level for the preregistered hypothesis was set at .05. Because we required all outcomes to be significant for the hypothesis to receive support, we did not adjust for multiple comparisons.
|Applicant orientation||Political affiliation of rater||Hireability||Willingness to cooperate and socialize||Competence||Warmth|
|Left||Rep||4.60 (1.65)||4.70 (1.53)||5.21 (1.07)||5.07 (1.19)|
|Dem||5.37 (1.21)||5.47 (1.09)||5.36 (0.94)||5.47 (0.93)|
|Right||Rep||5.23 (1.35)||5.45 (1.27)||5.55 (1.16)||5.51 (1.18)|
|Dem||4.23 (1.38)||3.59 (1.31)||4.83 (1.25)||4.23 (1.27)|
|Left||Rep||4.67 (1.58)||4.64 (1.56)||5.45 (0.94)||4.98 (1.37)|
|Dem||5.39 (1.14)||5.54 (0.95)||5.57 (0.88)||5.48 (0.98)|
|Right||Rep||5.44 (1.27)||5.64 (1.06)||5.80 (0.85)||5.51 (1.00)|
|Dem||4.35 (1.38)||3.78 (1.41)||5.02 (1.01)||4.23 (1.23)|
Note. All variables were measured on a scale from 1 to 7.
|2. Willingness to cooperate and socialize||0.660***||—|
***p < .001.
For the assistant, a 2 (applicant political orientation: Liberal vs. Conservative) × 2 (participant political affiliation: Democrat vs. Republican) between-subjects ANOVA showed no main effect of rater political affiliation, F(1, 475) = .85, p = .36, = .002, and a weak effect of applicant political orientation, F(1, 475) = 3.85, p = .05, = .008. Importantly, the hypothesized interaction effect was significant, F(1, 475) = 47.17, p < .001, = .09. The corresponding ANOVA for the IT-technician similarly revealed non-significant main effects for rater political affiliation, F(1, 490) = 2.39, p = .12, = .005, and applicant political orientation, F(1, 490) = 1.16, p = .28, = .002. Again, the expected interaction effect was significant, F(1, 490) = 55.53, p < .001, = .10 (see Figure 1).
Follow-up independent t-tests confirmed that Democrats gave higher hireability ratings to the liberal than the conservative applicant, both for the assistant, t(244) = 6.85, p < .001, two-tailed, mean difference = 1.13, 95% CI [.81, 1.46], Cohen’s d = .88 [.61, 1.14], and the IT-technician, t(263) = 6.68, p < .001, two-tailed, mean difference = 1.04, 95% CI [.73, 1.34], Cohen’s d = .82 [.57, 1.07].
Correspondingly, Republicans discriminated against the liberal (vs. conservative) applicant, t(231) = −3.19, p = .002, two-tailed, mean difference = −.63, 95% CI [−1.02, −.24], Cohen’s d = .42 [.16, .68] (assistant), and t(227) = −4.08, p < .001, two-tailed, mean difference = −.77, 95% CI [−1.15, −.40], Cohen’s d = .54 [.27, .80] (IT-technician). We observe that Democrats differed significantly more in their ratings of conservative and liberal candidates than Republicans did as the effect sizes for the Democrats were above the higher end of the 95% confidence intervals of the corresponding effect sizes for Republicans (see Cumming, 2009; standard confidence intervals based on the assumption of normality are reported here).
Willingness to Cooperate and Socialize
We ran the same type of ANOVAs and t-tests for the remaining dependent variables. For willingness to cooperate and socialize, there was a significant main effect of rater political affiliation for the assistant, F(1, 475) = 20.87, p < .001, = .04, and of applicant political orientation, F(1, 475) = 22.73, p < .001, = .05. These main effects were qualified by the hypothesized interaction effect, which was large, F(1, 475) = 121.18, p < .001, = .20.
Similarly, the ANOVA for the IT-technician revealed main effects of rater political affiliation, F(1, 490) = 17.94, p < .001, = .035, and applicant political orientation F(1, 490) = 11.12, p = .01, = .02. Importantly, they were qualified by a large interaction effect, F(1, 490) = 147.79, p < .001, = .23 (Figure 2).
The follow-up t-tests confirmed that Democrats gave higher ratings to the liberal than the conservative applicant. This was true for both for the assistant, t(244) = 12.26, p < .001, two-tailed, mean difference = 1.88, 95% CI [1.58, 2.18], Cohen’s d = 1.57 [1.28, 1.86], and the IT-technician, t(263) = 11.99, p < .001, two-tailed, mean difference = 1.76, 95% CI [1.47, 2.05], Cohen’s d = 1.48 [1.20, 1.45].
As expected, Republicans discriminated against the liberal relative to the conservative applicant, t(231) = −4.05, p < .001, two-tailed, mean difference = −.74, 95% CI [−1.10, −.38], Cohen’s d = .53 [.27, .79] (assistant), and t(227) = −5.70, p < .001, two-tailed, mean difference = −1.00, 95% CI [−1.35, −.66], Cohen’s d = .75 [.48, 1.02] (IT-technician). An inspection of the confidence intervals for the effect sizes reveals that the differences in ratings of conservative and liberal applicants were significantly and substantially larger for Democrats than for Republicans.
The ANOVA with competence judgments as the dependent variable revealed a significant albeit weak effect of rater political affiliation for the assistant, F(1, 475) = 8.07, p = .005, = .017, and no main effect of applicant orientation, F(1, 475) = .93, p = .34, = .002. Again, the interaction effect was significant, F(1, 475) = 18.28, p < .001, = .037.
Similarly, the ANOVA for the IT-technician revealed a main effect of rater political affiliation, F(1, 490) = 16.14, p < .001, = .03, no main effect of applicant orientation, F(1, 490) = 1.43, p = .23, = .003, and a significant interaction effect, F(1, 490) = 28.94, p < .001, = .056, see Figure 3.
Follow-up t-tests showed that Democrats perceived the liberal applicant as more competent, both for the assistant, t(244)= 3.79, p < .001, two-tailed, mean difference = .53, 95% CI [.26, .80], Cohen’s d = .49 [.23, .79], and the IT-technician, t(263)= 4.73, p < .001, two-tailed, mean difference = .55, 95% CI [.32, .78], Cohen’s d = .58 [.34, 83]. Correspondingly, Republicans perceived the conservative applicant as more competent, t(231) = −2.29, p = .02, two-tailed, mean difference = −.34, 95% CI [−.62, −.05], Cohen’s d = .30 [.04, .56] (assistant), and t(227) = −2.93, p = .004, two-tailed, mean difference = −.35, 95% CI [−.58, −.11], Cohen’s d = .39 [.13, .65] (IT-technician). In this case, the corresponding effect sizes for Democrats and Republicans were not significantly different.
The ANOVA with warmth judgments as the dependent variable revealed a main effect of rater political affiliation for the assistant, F(1, 475) = 18.16, p < .005, = .037, and a main effect of applicant orientation, F(1, 475) = 14.89, p < .001, = .03. The interaction effect was large, F(1, 475) = 65.05, p < .001, = .12. The ANOVA for the IT-technician also revealed a main effect for rater affiliation, F(1, 490) = 13.91, p < .001, = .028, of applicant orientation, F(1, 490) = 12.19, p = .001, = .024, and a large interaction effect, F(1, 490) = 73.26, p < .001, = .13 (Figure 4).
The follow-up t-tests confirmed that Democrats perceived the liberal applicant as warmer than the conservative applicant, both for the assistant, t(244) = 8.87, p < .001, two-tailed, mean difference = 1.25, 95% CI [.97, 1.52], Cohen’s d = 1.14 [.87, 1.41], and the IT-technician, t(263) = 9.17, p < .001, two-tailed, mean difference = 1.25, 95% CI [.98, 1.52], Cohen’s d = 1.13 [.87, 1.39]. They also confirmed that Republicans perceived the conservative applicant as warmer than the liberal applicant, t(231) = −2.83, p = .005, two-tailed, mean difference = −.44, 95% CI [−.75, −.13], Cohen’s d = .37 [.11, .63] (assistant), and t(227) = −3.33, p = .001, two-tailed, mean difference = −.53, 95% CI [−.84, −.22], Cohen’s d = .44 [.18, .70] (IT-technician). The confidence intervals reveal that the Democrats differed more than the Republicans in their ratings of conservative and liberal applicants.
We also conducted analyses that were not pre-registered. First, as Democrats and Republicans differed on average age and gender composition, we conducted sensitivity analyses, which confirmed that the size and statistical significance of the interaction effects in the ANOVAs remained virtually the same for all dependent variables when controlling for age and gender.
Second, we explored whether the interaction between rater political affiliation and applicant political orientation for the two dependent variables hireability judgments and willingness to cooperate and socialize could be explained by differences in stereotype content. We conducted mediation analyses using the PROCESS macro for SPSS (Hayes, 2022) with 10,000 bootstrap samples used to compute 95% confidence intervals. These revealed that the interaction effect of rater and applicant political affiliation on hireability judgments was (adjusting for the main effects) statistically mediated by both warmth (assistant: indirect effect = -.139 [-.193, -.091]; IT-technician: indirect effect = -.115 [-.159, -.078]) and competence (office assistant: indirect effect = -.078 [-.121, -.040]; IT-technician: indirect effect = -.111 [-.155, -.070]) judgments. The confidence intervals reveal that warmth mediated the interaction effect on hireability judgments significantly more strongly (p < .05) than competence in the case of the assistant. Similarly, the interaction effect of rater and applicant political affiliation on willingness to cooperate and socialize with the applicant was (adjusting for the main effects) substantially more strongly mediated by warmth (assistant: indirect effect = -.181 [-.240, -.126]; IT-technician: indirect effect = -.179 [-.234, -.130]) than competence (assistant: indirect effect = -.033 [-.061, -.011]; IT-technician: indirect effect = -.042 [-.071, -.019]) judgments. All variables were standardized prior to these analyses.
Finally, the explicit attitudes to cooperating and socializing with a politically similar vs. dissimilar co-worker were subjected to a 2 (Applicant orientation: Liberal vs. Conservative; within subjects) × 2 (Participant affiliation: Democrat vs. Republican; between-subjects) mixed ANOVA. There was a main effect of co-worker political orientation, F(1, 971) = 159.36, p < .001, = .14, and of rater political affiliation, F(1, 971) = 71.98, p < .001, = .07. The interaction effect was large, F(1, 971) = 722.53, p < .001, = .43. As expected, paired samples t-tests showed that Democrats would rather cooperate and socialize with a left-oriented co-worker (M = 5.51, SD = 1.05) than a right-oriented one (M = 3.81, SD = 1.40), t(510) = −25.51, p < .001, two-tailed, mean difference = −1.70, 95% CI [−.1.83, −.1.57]. Correspondingly, Republicans would rather do so with a right-oriented (M = 5.53, SD = 1.08) than a left-oriented co-worker (M = 4.92, SD = 1.39), t(461) = 11.68, p < .001, two-tailed, mean difference = −.61, 95% CI [.51, .72]. Democrats differentiated more between the two co-workers, compared to Republicans.
Considering the effort devoted to supporting diversity in organizations, the modest attention to the effects of political affiliation is surprising (Henderson & Jeong, 2022). Since many people are prejudiced against politically dissimilar others (Chambers et al., 2013), and most adults engage in daily interaction with other people at their workplace, it is important to study the consequences of political prejudice in work settings. Our findings clearly suggest that discrimination against applicants who reveal their political ideology is likely to occur, and its magnitude may not be trivial. This adds to the research by Gift and Gift (2015), who with a pooled sample of employers showed that applicants expressing a partisan view that differs from the majority have decreased chances to receive a job interview invitation.
Our findings further suggest that the social dimension – willingness to cooperate and socialize – is even more influenced by political orientation than judgments of hireability and competence are, confirming that partisanship is not only a political divide, but a social one (Iyengar & Westwood, 2015). If they manage to get hired in the first place, employees with a minority partisan view are at risk of becoming socially excluded at the workplace. Consistent with social identity theory and self-categorization theory (Turner et al., 1987), political orientation may become a basis for the development of in‐ and out‐groups at work, which could result in isolation for political minority employees (He et al., 2019). Many employees spend a substantial amount of their time cooperating with co-workers to achieve organizational goals, and the workplace often serves to fulfill social needs (Scott & Thau, 2013). Threatened belonging can thus lead to reduced social support (Beehr et al., 2000), diminished opportunities for self-worth enhancement (Crocker & Wolfe, 2001), and missed professional opportunities due to condensed social networks.
Regarding stereotype content, the interaction effect with political affiliation was particularly large for warmth, suggesting that political information may influence perceptions of an individual’s intent to either harm or help more than their capacity to act on their intentions (Fiske et al., 2002). For warmth, there was also a main effect of applicant orientation, suggesting that right-leaning applicants may be perceived as less warm overall. Further attesting to the importance of perceived warmth are the exploratory findings that although both warmth and competence significantly mediated the effects on willingness to cooperate and socialize, warmth was more important. Regarding hiring judgments, the results point to the possibility that warmth and competence are of equal importance for discrimination against politically dissimilar applicants in some occupations (IT-technician) whereas warmth is more important in other occupations (office assistant).
Although we observed discrimination and stereotyping among both Democrats and Republicans, Democrats were especially prone to exhibit bias in three out of four outcomes. For willingness to cooperate and socialize, and warmth ratings, this difference was particularly evident, suggesting that Democrats may be extra motivated to keep a social distance from their political opponents.
Many organizations lean either right or left; and ideologically skewed organizations tend to make ideologically consistent decisions (Gupta et al., 2017). Indeed, our findings indicate that applicants are at risk of having their employment prospects degraded when their political identities do not align with those of the people in the organization. Resume screening is often the first step in the selection process, and applicants who signal a political identity are therefore at risk of being rejected at this early stage. The most straightforward implication is that job applicants should avoid disclosing their political affiliation. Unlike attributes such as gender or age, political ideology is a deep-level characteristic that people can choose to either reveal or hide (Harrison et al., 1998). However, if certain belief systems are central to an individual’s identity, they may accidently reveal at least clues to their political ideology (Johnson & Roberto, 2019). Furthermore, hiding one’s political background by can be costly: Candidates whose primary credentials involve a political job or internship would have a hard time explaining the gaps in their resume.
Our results further suggest that once applicants are hired, they could become excluded from social collaborations or friendships at their new workplace. Furthermore, employees who have been at the workplace for a while will be exposed to more opportunities to discern each other’s political ideologies. The effects of political dissimilarity on employees’ chances to be socially included are therefore likely to accumulate over time.
For many individuals, political ideologies are as fundamental as religious beliefs (Jost & Amodio, 2012). Despite this, employees lack legal protection from being discriminated against based on their political ideology in many countries and in most U.S. states (Ballman, 2016). Our results suggest that policy makers may want to re-evaluate this matter. An additional reason is that political affiliation correlates strongly with legally protected group memberships such as race and religion (Johnson & Roberto, 2019). There is also a gender divide, where women are more likely to lean left. Hence, unfair treatment based on political affiliation can indirectly lead to discrimination (adverse impact) against legally protected groups.
Overall, the effects on discrimination and social exclusion could create increased polarization and decreased diversity in the organization, with possible long-term consequences for employees’ job satisfaction and productivity. Our findings thus point to the importance of including political beliefs and values in organizational diversity initiatives, with a focus on pluralistic diversity (i.e., inclusion of diverse views; He et al., 2019).
Strengths, Limitations, and Future Directions
Some of this study’s strengths include a theoretically and practically important research question, preregistered hypotheses and methods, and high statistical power. However, there are also some limitations. First, an extended experimental design including a baseline condition where political orientation is not mentioned could reveal more nuances, such as whether the discrimination effect is mainly driven by discrimination against outgroup members, or ingroup favoritism. However, in the correspondence study by Gift and Gift (2015), applicants with the same partisan affiliation as the majority were not more or less likely to receive a callback than non-partisan candidates, suggesting that it is primarily discrimination of those with opposing views that drives the effect. Second, we investigated discrimination and stereotyping in a context where applicants have equivalent credentials. A question for future research to pursue is whether rival partisans are disfavored even when they are more qualified. Treating applicants unfairly in this situation would perhaps be perceived as violating organizational justice (Cohen-Charash & Spector, 2001) which might override impulses to discriminate.
Regarding the ecological validity of the experimental manipulation, the applicant’s political orientation was signaled in the personal letter. Political orientation may also be signaled in the CV (see Gift & Gift, 2015) when describing previously or currently held positions of trust etc. Nevertheless, the fact that most participants noticed and remembered the political orientation signal, suggests that our manipulation worked as intended. Although it is unclear how common it is for job candidates to actually signal political orientation in a personal letter, there was nothing in our data to suggest that our candidate was perceived as odd. On the contrary, the relatively high (above the scale midpoint) overall ratings of hireability, warmth/competence, and willingness to socialize/cooperate suggest that our candidate was rather favorably perceived. An interesting extension, however, would be to investigate whether more subtle clues about political identity, such as being a gun owner or volunteering at a refugee center, might also produce discrimination.
Another limitation is that the applicant was always male, and although this makes it easier to compare the results to previous studies (Gift & Gift, 2015; Roth et al., 2020), it limits their generalizability. While it is unlikely that women would not face any discrimination due to their political orientation, it is of course possible that the amount of discrimination interacts with applicant gender. Future research should investigate this intriguing question, preferably while also randomly assigning different names (and other characteristics) to the applicants.
Furthermore, the results may not generalize equally well to other cultures. All countries differ in their political structure, and the meaning of the terms “conservative” and “liberal” differ by national context (Johnson & Roberto, 2019). In the US, the partisan divide is far-reaching and the political climate is charged. Although similar increases in the ideological divide between left and right have been observed in Western Europe (Silver, 2018), future studies could examine whether the effects replicate in countries with weaker tendencies of political polarization. It would further be interesting to examine this during election season, when political ideologies should be particularly salient.
We were interested in how employees would treat applicants and new co-workers: Not only executives and managers, but individuals throughout the organization may reinforce the dominating political ideology at the workplace (Gupta et al., 2017). Possibly, individuals working with personnel selection would differ in their judgments of politically active job applicants.
Our sample was diverse regarding age, gender, and education level (although women were in majority among Democrats). However, it consisted of individuals who had previously chosen to report their party affiliation when signing up to Prolific. It is possible that they have a relatively strong political identification. Nevertheless, the majority of the U.S. population identifies with or lean towards the Republican or Democratic Party (Jones, 2017), suggesting that the results are likely to generalize to most Americans.
Finally, the experiment was scenario-based and the findings may not generalize to real-life settings. For example, social desirability may have produced weaker effects despite the fact that our participants were anonymous. Ideally, this research should be replicated in a real hiring and workplace context. However, all participants held current employment and imagined their own workplace in the scenario, which should increase the probability that they perceived it as relevant.
In conclusion, our findings suggest that political prejudice can lead to both blatant discrimination and more subtle workplace mistreatment, making way for homogenous workplaces, with potentially far-reaching consequences for job prospects and interpersonal relationships at the workplace. Considering the expanding problems associated with political polarization, more research on consequences of political dissimilarity is warranted.