Firm Culture: Can Information Interventions Reduce Gender Gaps in Online Labor Markets?
How to cite
Annan, Francis; Archibong, Belinda; Okunogbe, Oyebola; Oliobi, Ifeatu. 2022. "Firm Culture: Can Information Interventions Reduce Gender Gaps in Online Labor Markets?." AEA RCT Registry. AEARCTR-0008841. https://doi.org/10.1257/rct.8841-1.1
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
Africa has both the world’s youngest population and some of the highest rates of youth unemployment globally. While a growing literature links gender and ethnic bias to inefficient economic outcomes across a range of contexts in Africa, there is limited understanding of their contribution to frictions in labor markets. Using new administrative data on 227,194 applicants and over 2 million matches from an online job platform in Nigeria, we study matching and the role of gender and ethnicity in firm hiring. We find significant differences in the matching behavior of applicants by gender. Women tend to be more qualified, by education, for the jobs they apply to than men, but less qualified, by years of experience. Women apply to lower numbers of jobs and more lower level jobs than men, even when qualified for the position. Women are also less likely to be hired for a position than similarly qualified male applicants.
Why do we observe these gender gaps in application and hiring? And can low cost information interventions help reduce gender gaps in applications and hiring in online labor markets? To answer these questions, we implement a randomized control trial (RCT) to examine the effects of low cost information interventions on the observed gender gaps in application and hiring in the online labor market.
Experimental design
Design 1: Effect of Information on Job Application Patterns on Type of Job Applied For.
Treatment: Message on gender gaps in job applications. Applicants will be randomly provided with 2 true statements using evidence from Jobberman's administrative data and then asked to answer if the statements are true or false. (a) Non-female information statement: “More than 50% of jobs are based in Lagos” and (b) Female information statement: “When equally qualified for a job, women are less likely to apply”. Applicants will be randomly provided with the true information about (a) and (b) (i.e., information treatment). Applicants then immediately prompted, upon receiving either information treatment, to choose to apply to one of 3 job levels: (1) Analyst, (2) Senior Analyst or (3) None.
Design 2(i): Effect of Female-Friendly Job Advert on Choice to Apply. We invite applicants to click on job description (from Design 1). For the respondents who choose 'Analyst" or "Senior Analyst" (from Design 1), they are randomly shown 2 different job ads. One describes generic duties of an entry-level part-time temp job with investigators at Barnard, and another which describes these details but includes one more detail that highlights the female-friendly nature of the position using the following prompt: “NOTE: We are committed to improving diversity in the research space and work to ensure that our team is supportive of our employees. We strongly encourage applications from women.” Importantly, the Analyst and Senior Analyst roles have the exact same job description as well.
Treatment: Job description includes a “female-friendly signals”: "We are committed to improving diversity in the research space"; “women encouraged to apply”; remote work, flexible hours" (all features of the job maintained but we randomly vary whether we include these signals or not). Our choice of signals is objectively determined by baseline survey evidence from the Jobberman's platform.
Design 2(ii): Effect of Female-Friendly Job Advert on Choice to Apply. This is similar to Design 2(i) but with listed firms on Jobberman's platform. We recruit 100+ firms willing to participate in the study to allow us to include a female-friendly message signals in their job ad.
Treatment: Among firms who agree, randomly select half to implement now (rest after 6 months). For treatment job ads, advertise to potential applicants via email campaign.
Design 3: Employer assessment of candidate profiles. We invite Hiring Managers to complete an online exercise to shortlist candidates for a given job description, who will then complete a short survey after the shortlisting exercise.
Treatment: Fictional candidate profiles to be rated and ranked (randomize candidate characteristics). We randomize job title (Analyst vs. Senior Analyst) across Hiring Managers.
Design details
Cited by
No citation data available for this registration yet.