Experimental Design
Discrete Choice Experiment with Garment Workers and Incentivized Resume Rating with Employers in Tamil Nadu:
Recruitment: I conduct a door-to-door listing survey in urban neighborhoods in Tirupur, Tamil Nadu, India located close to clusters of garment manufacturing factories. Survey teams are allocated to residential compounds and all households are surveyed within the compound. In each household, at most one person is surveyed (any adult member available at home), provided the household has at least one member who works in a garment factory. During the listing survey, I collect demographic information on all garment workers living in the household. Respondents are informed that as part of the study we will share worker profiles with HR managers at garment factories that are looking to hire new workers in the coming months. Respondents are asked to indicate whether they or anyone in their household would be interested in being recommended for work opportunities, and if yes, they are required to fill out a CV form to confirm their interest.
Sample: I use the data from the listing survey to identify married couples where both husband and wife are currently working in a garment factory or have worked in one during the past year. The sample includes couples currently working in the same establishment and those who work separately. I exclude households that said they were not interested in being recommended for work opportunities and those that did not fill out the CV.
Incentivized Job Choice Experiment: Once eligible households have been identified through the listing exercise, a pair of enumerators will return to the household to complete the main survey. Both husband and wife will be surveyed separately. The survey modules include: demographics, work history, beliefs about commuting safety and workplace safety, gender attitudes, spousal control/relationship quality. A hypothetical job choice experiment is also embedded into the survey. Before starting the job choice module, the respondent is informed that the research team is in contact with several HR managers at garment factories located close by who are looking to hire new workers and have randomly selected people from each household whose profiles will be shared with these HR managers. Respondents are told that they will be asked to choose between a series of hypothetical jobs and their responses will be used by the study team, to learn what kinds of jobs they like, so we can recommend them to opportunities that are aligned with their preferences. The respondent is also informed that at the end of the survey the enumerator will inform them who has been chosen from their household for the recommendation opportunity.
Each respondent is shown 7 pairs of hypothetical jobs with randomly varying wages and amenities, and asked to indicate which job they would prefer if they had the opportunity to do both. I randomly vary the following features of each job: (1) monthly salary (2) commute time (3) whether the job has a single vacancy or is hiring multiple people so both spouses could work together (4) whether the job entails night shift work (5) provisions for women's safety in the workplace (6) whether most of the co-workers can speak the respondent's native language. Randomization is at the individual x question level, and I stratify by gender, experience (<2 years or >= 2 years), and whether the respondent is currently working at the same establishment as their spouse or not.
Couples will be randomly assigned to one of two arms, stratifying by their current co-working status:
• “Joint recommendation” arm: both spouses will be recommended to HR, so they may potentially be hired together.
• “Single recommendation” arm: only one spouse will be recommended to HR; husband or wife randomly selected with equal probability.
In the weeks following the survey, couples will receive up to 4 phone calls informing them about different job profiles, identified based on their responses to the discrete choice survey. If the household decides that the profile of the selected spouse(s) should be shared with HR for a specific job opportunity, they must confirm their interest by calling a hotline number managed by the research team.
Incentivized Resume Rating Experiment: Eligible respondents will primarily be identified using the same listing exercise described above. However, since firm owners, supervisors, and labor contractors do not always live in the same neighborhoods as garment workers, we also rely on snowball sampling and directly approaching factories to identify eligible respondents. Subsequently, an enumerator visits the respondent to conduct the Incentivized Resume Rating survey. Before starting the resume rating module, the respondent is informed that the research team is conducting a survey of jobseekers interested in working at garment factories. Respondents are told that they will be asked to choose between a series of hypothetical candidates and their responses will be used by the study team, to learn what kinds of candidates are preferred at their factory, so we can recommend real worker profiles to them them that are aligned with their preferences. If they are interested, they can reach out to these real candidates, or refer them to their factory/HR departments, whenever vacancies arise. Respondents are shown a series of pairs of hypothetical worker profiles (5 male and 5 female), with randomly varying features. I randomly vary the work experience of each hypothetical jobseeker and the type of referral relationship. For female candidates, I randomize whether they were: not referred by anyone, referred by their spouse, or referred by their female friend. For male candidates, I randomize whom they are likely to refer in the future: no one, their spouse, or their male friend. Respondents must then answer a series of questions about each profile: (1) which applicant they would prefer to hire (2) how long they expect the applicant to stay if they are offered the job, (3) whether they would assign the worker to the same team as their referrer (4) what wage they would offer to the applicant, (5) whether they would hire the applicant in an entry-level (helper) role or the senior role, and (6) whether the referrer is likely to interfere if the supervisor is reprimanding the candidate for production errors.
Discrete Choice Experiment with Married Couples in Odisha:
Recruitment: I conduct door-to-door recruitment in villages located within the catchment area of a new industrial park being constructed in Khordha district, Odisha. Several garment factories are expected to begin operations in this park in 2026. Enumerator teams are allocated to hamlets and will target all households in a given hamlet. After approaching a house, enumerators ask to speak to a married woman aged between 18-60, and ask questions to determine eligibilty: whether the woman lives with her husband/husband’s family, whether she does more than 20 hours of paid work per week in a typical month during the year, and whether she is interested in new work opportunities. Enumerators ask the respondent if they are interested in new work opportunities using a 4 point Likert scale (not at all interested, not interested, interested, very interested). If the woman reports that she is does not live with her husband or husband’s family, engages in more than 20 hours of paid work per week, or is not at all interested in work opportunities, the household is screened out of the study sample. Husbands are only surveyed if their wives satisfy these eligibility criteria.
Experimental Design: Respondents are informed that we are conducting a study to understand people's job preferences. Each spouse is surveyed separately and asked to choose between a series of pairs of hypothetical jobs. Women are asked to imagine their husband is employed at a garment factory and choose between two jobs for themselves: one at the same factory and another at a different factory in the same industrial park. Job attributes, including wage and whether a female friend or relative also works at the factory, are randomly varied. Men are asked to imagine they are working at one of the garment factories, and similarly asked to choose between jobs for their wives. Respondents are asked to imagine that the two jobs are the same in all other ways, and must indicate which they would prefer. Women must report whether they would accept the offered job, while men are asked to report whether their wife should accept the offered job.
To understand how co-working shapes labor supply responses to local jobs and migration opportunities, respondents complete two sets of choice tasks: one for local jobs in the new industrial park near their village (5 pairs of jobs) and one for jobs in a distant industrial park being built 300 km away (3 pairs of jobs).
Finally, participants select whether they wish to receive hiring updates via text for one of five job types (jobs for both husband and wife at the same factory, jobs for the husband and wife at separate factories, jobs for the husband alone, jobs for the wife alone, or none of the above), providing an incentivized measure of job preferences.