Experimental Design
We will create resumes of fictitious candidates for each of several different job titles, working closely with our partner to ensure the fictitious resumes are realistic. We will invite job recruiters connected to our partner to evaluate the hypothetical resumes in an online (qualtrics) survey. This will be done through Incentivized Resume Rating (IRR): recruiters will know the resumes are fictitious, but we will match them with real candidates based on their answers, which incentivizes them to carefully and honestly rate the resumes.
We will randomly vary key elements of the resumes to estimate recruiter preferences over various candidate characteristics. More specifically, each recruiter will rate 43 resumes of candidates in their field. The first resume will be the same for all recruiters looking to fill a given position and will present a “neutral” candidate (a male with no career break and average qualifications); this initial resume is meant to reduce priming effects and evaluations of it will be excluded from our analyses. Each of the subsequent 42 resumes will be based on an underlying template with key details of the resume randomized across employers. The templates will be based on real resumes our partner shared with us and will all reflect candidates with an undergraduate degree and two work experiences. The order in which the 42 templates appear will be randomized across recruiters.
We will randomize the following elements on top of the templates: the candidate’s name (which embeds gender and caste), years since completion of undergraduate degree, the prestige of the undergraduate institution, the undergraduate grade percentage, whether the candidate has a masters degree, the duration of the more recent work experience, whether the more recent employer is a prestigious employer, whether the skills the candidate lists are basic or advanced, whether the candidate has a career break, and details of the career break. The break details to be randomized depend on the gender of the candidate. For women, we will randomize whether the break is “resolved” (i.e. happened before the more recent work experience) or “unresolved” (i.e. there has been no work experience since the break), and whether the woman explicitly calls the break a maternity break or says nothing but has a gap in her experiences. Among the women who label the break as a maternity break, we will further randomize whether women say the break was to care for two daughters or a daughter and a son, and whether women did any “upskilling” (i.e. some sort of course or training) during the break. For men, we will randomize whether the break is resolved or unresolved, but nothing else; any breaks on male resumes will not be labeled and will simply be a gap in experiences. Additional details of the randomization are attached in a supporting document.
We will ask recruiters to complete a brief survey after rating the resumes that asks about recruiter demographics and characteristics of their firms, among other topics.