What Works for Her?: How Work-from-Home Digital Jobs Affect Female Labor Force Participation

In this study, we use novel jobs that are normally performed by Business Process Outsourcing (BPO) units on Personal Computers/Desktops and convert them to smartphone-based micro-tasks that can be performed by women from home. These tasks are presented on a tailor-made Android mobile application, called Rani (Queen). This is a gig economy platform allowing for flexible work hours. They are all shown a similar set of tasks, their work is validated for accuracy, and payments are made weekly.

The intervention is about 50-100 hours of work per worker (or 1-1.5 months) on this digital job platform for women over their own smartphones. The treatment arms differ in the job contract offered to the women, specifically by location of job and payment level offered.

In this study, we use novel jobs that are normally performed by Business Process Outsourcing (BPO) units on Personal Computers/Desktops and convert them to smartphone-based micro-tasks that can be performed by women from home. The job task of image classification is standardized using existing recommendations, along with text classification and sentiment analysis tasks. These tasks are presented on a tailor-made Android mobile application, called Rani (Queen). While available on Playstore, only the workers assigned in the experiment can earn income on it. Workers sign up with their phone number and are immediately shown their job task. This is a gig economy platform allowing for flexible work hours, so women can log in whenever they would like and work for however long they want to and work for however long they want to, with a recommendation to work for 2 hours per day. They can bring their children along to the center if they would like to as well. They are all shown a similar set of tasks, their work is validated for accuracy, and payments are made weekly.

The intervention is about 50-100 hours of work per worker (or 1-1.5 months) on this digital job platform for women over their own smartphones. The treatment arms differ in the job contract offered to the women, specifically by location of job and payment level offered.

2022-09-28

2023-05-31

1. The primary outcomes through the baseline survey are the job take-up results that are measured (1) at the end of the survey, (2) through downloading the job application, and then (3) through observed work completed on the application. The job acceptance results are compared across all treatment arms (home vs center, and the three payment levels, and knowledge of husband survey).
2. We also learn about women’s willingness to work, and to take on future work engagements, along with willingness to switch from WfC to WfH (at lower pay) or from WfH to WfC (at higher pay) at Endline.
3. The study also tries to understand the effects of these digital job arrangements on women’s mental health, agency, empowerment, and other well-being indicators.

Primary outcomes -
Job take-up rates are measured by observing whether women are actually completing tasks on the Rani platform (more than 0 tasks, and more than 10 tasks) right after baseline survey. We also measure job take-up during the two baseline surveys where we ask women to self-report whether they are accepting or rejecting the job.
Mental health indicators are measured using the PHQ-9 scale (using the first 8 measures), and tension/stress scale developed for this context measured at baseline, and endline (after 2-3 months from baseline).
Agency and empowerment outcomes are measured using a combination of indices looking at: (all measured at baseline and then at endline 2-3 months from baseline) using various likert scales
(1) women's agency and decision making at home (within household measures, and being able to pursue career of own choice, restrictions in ability to work). These also include measures around women's estimated economic value for how much women should be paid for doing housework in their own house (economic value of a woman, as measured by women and their husbands), and domestic violence metrics (whether a woman believes if she goes against her husband, then he will shout at or beat her, and whether domestic violence increases/decreases when a woman first starts to earn money)
(2) financial independence measured by having own bank account under self name, and savings
(3) mobility (being allowed to visit places using public transport)
(4) social network (number of women on the floor/building who she regularly speaks to)
(5) confidence measured as level of confidence in being able to do a job
(6) aspirations measured by aspirational income level and how different it is from their actual income level

1. We measure intensive margin outcomes for women’s work, including the amount of time spent on the application, number of tasks completed, accuracy levels (measures their job productivity), and total payment earned.
2. For heterogeneity analysis, we explore the attributes of women who are more or less likely to accept job offers across the different treatment arms.
3. Husband’s perceptions and decisions: We measure the role of husbands in the decision-making process through husband surveys, and by informing some women their husbands would be surveyed
4. Time use: We understand how women are spending their time in household chores and how that might change because of the job intervention.

Secondary outcomes -
Job retention rates are measured by number of women in each treatment arm who work for >500 tasks, >5,000 tasks, >10,000 tasks, and completed all 15,000 tasks on the platform during the job intervention duration.
Social norms: (all measured at baseline and then at endline 2-3 months from baseline) using a likert scale
(1) Changes in perceptions of women working/not working, perceptions of others' perceptions of women's work, changes in decision to work after not being able to work
(2) Dignity: where woman's self-worth and family's image lie in a woman following social norms
(3) Capabilities: whether a woman thinks she can manage a household and a job together
(4) Insecurities: whether a woman would be worried if she started to earn more than her husband
Time-use is measured at baseline and then at endline 2-3 months from baseline where women are asked how their spend the different hours of the day working, and doing other household chores during the day.

Women are assigned to one of seven arms (6 treatment, and 1 control). The 6 treatment arms vary in the job location (work-from-home or WfH, work-from-center or WfC) and payment level (low, medium, high). Baseline surveys are completed with all eligible women in the study. At the end of the survey, women are offered a job contract. The job contract includes details about the job task, payment details, and the location of the job. Around 20% of the women are also told that this job contract information will be informed to their husbands through another survey conducted by a male enumerator.

Women are then asked whether they accept the job contract. After the job contract offer, women do the job for about 50-100 hours (over ~1.5 months) before the Endline. At Endline, women are offered different job contracts, switching the location, and adding other variables of distance to center, among other job characteristics.

At endline, specifically we have 5 treatment arms - (1) work from home, (2) work from home but with a visit to a local center, (3) work from home but with no multitasking at home, (4) work from home but with a visit to a local center and wear a worker ID, and (5) work from local center to understand the mechanisms behind any differences between home and center job take up. We also run an experiment with husbands, offering them the same job offers as their wives (low, medium, high wages from both locations - home and center) to elicit their preferences and job take-up for a short duration for comparison with wives' responses.

Women are assigned to one of seven arms (6 treatment, and 1 control). The 6 treatment arms vary in the job location (work-from-home or WfH, work-from-center or WfC) and payment level (low, medium, high). All the centers in the WfC arms are within the local community, where women can walk to their assigned one in 2-10 minutes. So, many of the usual “outside costs” of women's work are either limited or do not exist. Hence, we see the WfC as a “social signal” of the woman's work, while WfH is private.

Baseline surveys are completed with all eligible women in the study. At the end of the survey, women are offered a job contract. The job contract includes details about the job task, payment details, and the location of the job. Around 50% of the women are also told that this job contract information will be informed to their husbands through another survey conducted by a male enumerator. These husband survey offers are offered to a random subset of households within each arm. A random subset of the control arm also receives the husband survey. For the remaining households, they are not told anything about a husband survey. Women are then asked whether they accept the job contract.

After the job contract offer, women do the job for about 50-100 hours (over ~1.5 months) before the Endline. At Endline, women are offered different job contracts, switching the location, and adding other variables of distance to center, among other job characteristics. Women are also offered more full-time work and we observe the types of women who are most agreeable to other types of jobs across the treatment arms. At endline, specifically we have 5 treatment arms - (1) work from home, (2) work from home but with a visit to a local center (inconvenience + limited social signaling), (3) work from home but with no multitasking at home (no multitasking), (4) work from home but with a visit to a local center and wear a worker ID (inconvenience + more social signaling), and (5) work from local center to understand the mechanisms behind any differences between home and center job take up. We also run an experiment with husbands, offering them the same job offers as their wives (low, medium, high wages from both locations - home and center) to elicit their preferences and job take-up for a short duration for comparison with wives' responses.

Randomization done by a computer.

Individual households.

No

NA

3000-4000 households

Around 500-530 per treatment arm (treatments 1, 2, 3, 4, 5, 6, 7) and around 500-530 for control.