The impact of religious identity on hiring outcomes

This study consists of a factorial survey experiment in which recruiters evaluate fictitious job candidates. Candidate profiles are presented in the format of an HR software interface and simulate the first stage of a recruitment process. Each respondent evaluates four candidate profiles and indicates hiring-related evaluations and perceptions.

Candidate profiles vary systematically across several experimentally manipulated dimensions, including gender, religious affiliation, visual signals of religion, ethnic background signalled through names, and extracurricular activities. The purpose of the intervention is to measure how religious identity and related signals affect hiring decisions and employer perceptions.

Profile pictures used in the experiment were selected based on a separate pre-registered photo validation study (https://www.socialscienceregistry.org/trials/18099).

This study employs a factorial vignette experiment to examine hiring discrimination based on religious identity and related ethnic signals. Recruiters evaluate fictitious school-leaving candidates who apply for jobs within the respondent’s own company.

The experimental design varies five candidate dimensions:

1. Gender (male; female)
2. Visual signal of religion (present; not present x2)
3. Ethnic background signalled through names (minority; majority x2)
4. Extracurricular activities (religious volunteering; general volunteering x2)
5. Religion (Islam; Christianity x2)

Candidate profiles are presented in a realistic HR software interface. Respondents evaluate four candidates and provide assessments of hiring probability and their perceptions regarding the candidate.

Profile photographs were selected using a separate pre-registered photo pretest. Because it was not feasible to obtain a sufficient number of photographs that were statistically equivalent within a narrow equivalence margin around the midpoint of perceived ethnic ambiguity, photographs were selected based on proximity to the midpoint of perceived ethnic origin. In the analysis, photograph fixed effects will be included in all regression analyses to control for all non-manipulated photograph characteristics, including perceived ethnicity and attractiveness.

The order of vignettes and survey items is randomized. The study also includes indirect questioning techniques and a Bayesian Truth Serum procedure to reduce social desirability bias.

2026-04-19

2026-05-19

- Hiring probability: likelihood that the candidate will be invited to the next stage of the hiring process (11-point Likert scale).
- Taste-based discrimination perceptions: expected willingness of employer, colleagues, and clients to work with the candidate (11-point Likert scale).
- Statistical discrimination perceptions: 6 HEXACO dimensions, candidate personality traits: honesty–humility, emotionality, extraversion, agreeableness, conscientiousness, openness (11-point Likert scale).

For taste-based discrimination and statistical discrimination, composite indices are constructed by taking the mean of the items within each conceptual cluster. Item-level analyses will also be conducted.

- Perceived religiosity of the candidate (11-point Likert scale).
- Perceived religious affiliation of the candidate (Christian, Muslim, atheist, or unknown)

This study is a factorial survey experiment in which recruiters evaluate fictitious job candidates. The experimental design varies five candidate characteristics: gender, religion, visual religious signals, ethnic background signalled through names, and extracurricular activities.

The full factorial design results in 162 unique candidate profiles. These profiles are distributed across 41 vignette decks, each containing four candidate profiles. Each respondent is randomly assigned to one deck and evaluates four candidates. The order of candidate profiles is randomised.

The experiment simulates a realistic hiring context in which recruiters evaluate candidates for positions within their own company.

This study employs a factorial survey experiment to examine hiring discrimination based on religious and minority-associated ethnic identity markers. The experimental design systematically varies five dimensions: gender (male; female), visual signal religion (not present; present x2), ethnic background through name (minority; majority x2), extracurricular activities (religious volunteering; general volunteering x2), and religion (Islam; Christianity x2).

The different dimensions and accompanying levels are operationalised as follows:
- Gender: signalled through names and by explicit mention in the candidate profiles.
- Visual signal religion: hijab for Muslim women, beard for Muslim men, cross pendant for Christian women and men.
- Ethnic background through name: Turkish names for Muslim ethnic minorities, Polish names for Christian ethnic minorities. Specific names extracted from Martiniello & Verhaeghe (2022).
- Extracurricular activities: volunteer at church/mosque for religious volunteering, volunteer at local community centre for non-religious volunteering.
- Religion: this dimension determines the specific interpretation of the above dimensions. If religion or ethnic minority is not indicated anywhere in the above dimensions, this will not be reflected in the candidate profile.

This 2 × 3 × 3 × 3 × 3 design yields a vignette universe of 162 unique combinations. We generate 164 combinations using the %MktEx macro in SAS and partition these into 41 decks, each containing four vignettes, using the %MktBlock macro.

As described in the study design, the %MktBlock macro is first used to allocate the vignettes to decks.

Participants are randomly assigned to one of the decks through Qualtrics. Within each deck, the order of vignette presentation is randomised through Qualtrics to mitigate potential order effects.

Respondents are randomly assigned to vignette decks, and vignette order is randomized within respondents.

No

205 recruiters, each recruiter evaluates 4 candidate profiles.

Total expected observations: approximately 820 vignette evaluations

Treatment variation occurs at the vignette level within respondents.

The sample size is determined by the factorial survey design. Following Auspurg and Hinz (2014), each vignette deck should be evaluated multiple times to obtain stable estimates. With 41 decks and a target of at least five evaluations per deck, the required sample size is 205 respondents.

Analysis plan