Bringing Work Home: Internet-Mediated Gig Work and Women's Employment

The intervention consists of month-long internet-mediated gig jobs which can be done on a smartphone. Participants complete speech-based tasks on an Android application developed by Karya Inc. We pay participants a piece rate of approximately Rs. 1 per task, which is the wage our implementing partners have previously used in similar contexts. These tasks include recording oneself speaking specific sentences in Bengali or Hindi, and the tasks range in difficulty level.

Participants are randomized into work arrangements which vary across three characteristics: (1) the ability to work from home, (2) the ability to choose one's work hours flexibly, and (3) the ability to multitask with childcare. After making job offers but before the work begins, we randomly select half of participants in the treatment groups with inflexible constraints to be upgraded to the most flexible work arrangement, which allows them to work from home, choose work hours, and multitask with childcare.

2022-06-17

2022-09-17

1. Interest and take up of jobs by work arrangement. The work arrangements vary across three attributes: (1) ability to choose location, (2) ability to choose hours of work, and (3) ability to multitask with childcare. We will assess interest/take up of jobs in at least two ways: (1) women's responses in the baseline survey to whether they would take each job if offered it, and (2) whether or not women actually start the job that they are assigned during the intervention period.

2. Interest in further training and work after the intervention. During the endline survey, we will elicit women's interest in skills training or other work opportunities. We then offer another round of work opportunities to all participants who completed the endline survey. These work opportunities vary in their flexibility in time, ability to multitask with childcare, and work location. We measure take up of these job outcomes.

3. Gender attitudes related to gender and work, household roles, and technology.

4. Women's agency. We will measure agency related to at least two categories: (i) mobility and (ii) ability to make financial decisions.

1. Attributes of women who select into different work arrangements. We will measure whether women who opt into more or less flexible jobs differ along (i) baseline characteristics such as agency, gender attitudes, and household structure, (ii) on-the-job productivity, and (iii) persistence in the job throughout the month-long intervention.

2. On-the-job productivity, including comparisons across work arrangements.

3. Persistence in the job (i.e. to what extent women continue in the job throughout the intervention), including comparisons across work arrangements.

4. Children's aspirations and attitudes. We will compare the stated educational/career aspirations and gender attitudes at endline between children in treatment households vs control households. We do not expect to be well-powered on this outcome.

5. Time use. We will try to understand how women change their schedules in order to accommodate work in the treatment group.

6. Spending. We will try to understand how households use the earnings that women make from the intervention.

Our evaluation consists of an RCT in and around Kolkata, West Bengal with approximately 1,500-2,000 households. Participating households will be randomly assigned into job treatment groups and a control group. The month-long, smartphone-based jobs consist of contributing to a Bengali or Hindi dataset by speaking aloud provided sentences.

The job treatment groups vary across three dimensions: 1) ability to choose work location, 2) ability to choose work hours, and 3) ability to multitask with childcare. First, we offer jobs to women and measure the difference in job acceptance for each job variation. Second, we return to a random subset of participants who were originally offered inflexible work arrangements, and we offer them the option to switch to the most flexible job. Third, we implement the month-long jobs. Lastly, at the end of the short-term job, we provide information about further training and work and measure the impacts of jobs on the households (e.g. women's interest in future work, time use, gender attitudes, and agency). To measure the impact on women's interest in future work, we offer another round of jobs and measure take up of these job opportunities. These jobs vary in whether they are digital or non-digital, and vary across the same three dimensions as the original treatment.

Randomization done by a computer

Household

No

N/A

1500-2000 households

1250 treatment, 420 control;

First job offers: 208 most flexible job, 208 fixed time-slot job, 208 for no multitasking with childcare, 208 fixed time-slot and no multitasking with childcare, 415 fixed location.

Final job offers (after option to switch to most flexible job): 731 most flexible job, 103 fixed time-slot job, 105 for no multitasking with childcare, 105 fixed time-slot and no multitasking with childcare, 206 fixed location.