Our research agenda addresses questions central to the economics of gender: To what extent are gender gaps in labour markets driven by labour demand versus supply? Given weak capacity in developing countries to enforce legislation, can market-based interventions help redirect the search of workers and firms towards more equitable matches?
We leverage our research partnership with Pakistan’s biggest online job search platform (Rozee.pk) to make progress in answering these questions. On Rozee.pk, we observe comprehensive information on job seekers, including current and desired wages, work experience, and application choice sets. For a given job, we also observe applicant pools, salary ranges (even if not revealed to workers), and detailed job descriptions. Using text analysis, we gather more data on the non-wage amenities and task complexity of jobs. With these data, we are able to follow the full trajectory of workers from their applications to recruitment outcomes and document gender gaps at each stage of the job search process. We then decompose the observed gender gaps into demand and supply-side factors. In addition, we test two policy interventions:
- We study the effect of salary range disclosure in job ads on candidates’ sorting, bargaining, and hiring outcomes. In particular, we study effects on the gender pay gap as informational frictions regarding pay may affect women more adversely than men.
- We investigate why many firms in developing countries (19% on Rozee.pk) explicitly state a preference in job ads for hiring men and whether they reconsider this when exposed to debiasing information about relevant female job seekers in the market.
Existing studies on gender gaps in labour markets are typically done in partial equilibrium, holding labour supply or demand fixed. Alternatively, they are based on equilibrium relationships observed in employer-employee matched data, which require strong assumptions for inference as they do not observe choice sets, and these data are rare to find in developing countries. Recent studies using online job platforms have attempted to address these issues, but they are typically based in high-income countries and rely on specialized job boards, e.g., industry- or firm-specific platforms, or platforms designed for unemployed workers. With 95,000 firms, 6 million job seekers, and 45 million applications, Rozee.pk receives more visitors than the next 5 largest job sites in Pakistan combined. As a result, our study offers unparalleled coverage of jobs and worker-firm interactions across a vast range of occupations and industries in Pakistan.
We also provide the first experimental evidence on the effects of salary range disclosures in job ads, which have just been enacted in four US states and piloted in the UK. Three recent papers have studied related reforms. Skoda (2022) and Frimmel et al. (2022) find limited effects of revealing the minimum salary in a job ad on the gender pay gap. But our proposal to reveal the full range is crucially different since inferring a firm’s max willingness to pay is often the bigger challenge in wage negotiations. Arnold et al. (2022) study the effects of Colorado’s mandate to disclose the salary range in job ads, but only on advertised (not actual) salaries and through spillover, rather than direct impacts. They are unable to consider effects on labour supply, strategic firm responses, and the composition of non-compliers. Our research is distinct because we randomize the disclosure of salary ranges, study effects on actual wages as well as demand and supply responses, and also account for selection into compliance since we observe the salary range of a job even if it is not disclosed in the job ad.
Finally, while it has been shown that in China, exogenously removing stated gender preferences from job ads led to more female applicants and hires for historically male jobs (Chen and Kuhn, 2022), little is known about why firms state these preferences and how to convince (rather than coerce) them to remove them. In a context where law enforcement is low, finding a self-enforcing, market-based solution is crucial.
We use a discrete choice framework as in McFadden (1978) to document gender gaps at each stage of the job search process by modelling the decisions of firms and workers throughout the hiring process. Additionally, we conduct firm surveys with embedded hypothetical job ad vignettes to capture firm reasoning for (i) concealing the salary range and (ii) stating gender preferences in job ads. We then use randomized experiments on Rozee.pk to study the effects of our proposed policy experiments on a vast range of gender-specific sorting, bargaining, and hiring outcomes.