We study the interlinked role of beliefs, preferences, and search by workers and firms in driving occupational segregation and wage inequality for women in Nairobi, Kenya. Our project address four questions within this agenda:
(1) (How) do employers treat women and men differently in the hiring process? (2) (How) do men’s and women’s preferences over job attributes, beliefs about the labour market, labour market choices, and search strategies differ? (3) (How) does providing accurate information to workers about the attributes employers value and their recruitment strategies affect men’s and women’s search strategies? (4) What portion of gender segregation can be accounted for by differences in: job seeker preferences, employer preferences, job seeker misperceptions, and search strategies?
Our project is motivated by pervasive gender gaps and labour force participation across the developing world. Women in Kenya are more likely to work in the informal sector and unpaid for family, and less likely to work in wage employment. Within wage work, women are more likely to have shorter tenure and more likely to work in low productivity sectors. We hypothesize that these differences may be the product of both supply- and demand-factors. For example, they might reflect women’s preferences for non-wage job amenities, beliefs about how employers will evaluate them, and the methods and networks they can access when searching. On the demand side, firms might discriminate against women or believe that women are more likely to quit due to childcare responsibilities. Understanding the relative importance of these factors through observational data or a study that examines a single constraint at once is challenging—the equilibrium allocation of workers to jobs reflects these factors operating in tandem. This interlinked nature motivates our plans to collect detailed data on preferences, beliefs, and search methods from both sides of the labour market.
We will address our research questions through five activities:
(1) a firm survey in which we ask about hiring practices, and a CV rating task, in which we measure firm’s preferences over jobseekers’ attributes, including their gender and parenthood status, and beliefs about jobseekers’ response to offers; (2) a jobseeker survey in which we measure female and male jobseekers’ preferences over job attributes, and beliefs about job search outcomes (at baseline and endline) (3) an information treatment, in which we use information from our firm survey to experimentally vary jobseekers’ information sets about the attributes valued by firms across sectors, and how sectors recruit workers; (4) a high-frequency panel phone survey of jobseekers, in which we measure gender differences in search, application and work behavior; and (5) a structural model of job search and hiring, estimated using the data we collect, to quantify the relative importance of beliefs, preferences, and search on the worker and firm side to gender gaps and occupational segregation.
We will first recruit 120 firms across 12 sectors. These sectors will vary in their gender composition and competitiveness, but will share the property of requiring a secondary education. We will construct hypothetical CVs of jobseekers that vary in their gender, education, qualifications, and other attributes. We will also vary whether the applicant is a parent, and whether they were referred by someone at the company. Firms will answer questions about their recruitment, hiring, and employment practices, and will rate these hypothetical CVs, on their interest in hiring the candidate, how likely they think the candidate would be to accept the job, and how likely the candidate would be to remain at the company a year later.
We will then conduct a listing of young women and men in two settlements of Nairobi. We anticipate listing 6,000 households to enroll 1,600 adults aged 18-35 with a secondary degree or higher who are actively searching for work. They will be asked to participate in a baseline survey, where we measure their preference over various job attributes, such as flexibility, commuting time, and formality, and measure their beliefs about their competitiveness in different jobs. We will also ask respondents to evaluate a subset of the same hypothetical CVs. Individuals in our treatment condition will learn the ratings that firms provided, and comments they offered about CVs. They will also receive access to “report cards” about the skills valued by different sectors, their competitiveness, and the methods used by these firms to recruit.
Following the baseline, we will conduct a high-frequency panel phone survey with these young women and men. This will enable us to observe short-term search strategies and work that longer-run surveys plausibly miss. We will observe applications submitted, expectations about future arrivals, job offers received, labour supply, and job separations. Finally, we will conduct an endline survey of the individuals, to observe how their preferences and beliefs have evolved over time, both in response to the treatment, and due to the realizations of their search effort and work experiences.
We will use these data to build and estimate a dynamic model of workers’ search behavior and occupational choice which allows for rich preferences and potentially biased beliefs. We will strategically design our survey instruments to allow us to identify the structural parameters of our model transparently and credibly. We will use this model to explore channels that drive occupational segregation by gender and gender gaps which we cannot experimentally identify. For instance, how much would gender segregation change if we equalized the “true” demand-side treatment of young men and women in the labour market?
Our research will make several contributions to the existing literature on labour markets in developing countries, where data on women and men’s beliefs and preferences on either side of the labour market is quite rare. To the best of our knowledge, this will be the first study to elicit both rich preferences and beliefs from both sides of the labour market. Collecting such data, and interpreting it through our structural model, will enable us to quantify the role of each component in driving gender inequality. Second, our work will speak to the external validity of findings on female worker preferences in higher-income countries, such as a demand for flexibility, shorter commutes, and the ability to engage in part-time work. Third, on the firm side, only a handful of studies have measured either firm beliefs or preferences over workers in developing countries, with limited attention to how gender and fertility shape firm preferences. Finally, our information intervention will complement existing research on the role of beliefs in job search in low-income countries. While interventions such as transport studies and job matching appear to impact behavior by altering beliefs, we will directly provide job seekers with fairly detailed, accurate information, and examine how this information affects the specific mechanics of search and the evolution of belief-updating over time in a high-frequency panel.