Working Papers

What Works for Her? How Work-from-Home Digital Jobs Affect Female Labor Force Participation in Urban India (with Lisa Ho) 

In many developing countries, married women face significant barriers to entering the workforce, often rooted in gender norms. These may manifest as practical constraints—like travel restrictions and housework responsibilities—and other intangible domesticity constraints like household disapproval of women working outside the home. We design an experiment to distinguish between these barriers by establishing new job offices for part-time, smartphone-based digital work with minimal practical constraints: the work is local, only for women, less than five minutes walking distance, and it permits children. We assigned 3,200 wives in Mumbai to work-from-home or office jobs and cross-randomized them to one of three monthly wage levels (low, medium, or high). We find that 56% of wives started working from home, while only 27% took up office jobs, matching India’s female labor force participation rate. Surprisingly, wages doubling household income did not significantly affect entry. A parallel experiment with husbands showed more responsiveness to wages and no preference for the job location. A follow-up experiment to uncover the underlying mechanisms finds that introducing an observable two-minute daily check-in at the office decreased home job take-up by 25% (explaining half the home-office difference), driven by women from less progressive households. Taken together, the experiments suggest that even beyond practical constraints, domesticity constraints—the belief that a woman’s place is at home—restrict women's labor market entry in India. Without changes to these constraints, home-based jobs may represent the most immediate path to increase women's labor force participation.

Featured in: 

World Bank Development Blog

Stanford Hoover Institute Economics Applied Podcast

Substack Blog Post / Podcast

Bringing Work Home: Flexible Work Arrangements as Gateway Jobs for Women in West Bengal (with Lisa Ho and Anahita Karandikar)

Several hundred million women want a job but are out of the labor force, often because available opportunities are incompatible with traditional norms about their household roles. In a field experiment with 1,670 households in West Bengal, we offer flexible, short-term data entry jobs which meet households where they are in terms of expectations on women’s domestic responsibilities. We find three sets of results. First, flexibility more than triples job take up, from 15% for an office job to 48% for a job that women can do from home, while multitasking with childcare, and at the hours they choose. Second, taking the perspective of a firm offering flexible work arrangements, flexibility has no adverse effects on the amount of quality-adjusted output that workers produce, but workers are less efficient from home. Third, flexible jobs act as a labor market gateway for women initially out of the labor force: experience with flexible jobs makes women more likely to accept less flexible and outside-the- home jobs in the future. This gateway effect may be explained by changes in attitudes about appropriate behavior for men and women. Flexibility makes a larger difference to the labor supply of women who hold more traditional pre-intervention attitudes, and work experience in turn shifts women and children’s gender attitudes to become less traditional. Thus, flexible work arrangements can both attract women to the labor force and provide a gateway to less flexible jobs.

Publications

Shrinivas, A., Jalota, S., Mahajan, A., & Miller, G. (2023). The Importance of Wage Loss in the Financial Burden of Illness: Longitudinal Evidence from India. Social Science & Medicine, 317. https://doi.org/https://doi.org/10.1016/j.socscimed.2022.115583

Abstract: A key aim of Universal Health Coverage (UHC) is to protect individuals and households against the financial risk of illness. Large-scale health insurance expansions are therefore a central focus of the UHC agenda. Importantly, however, health insurance does not protect against a key dimension of financial risk associated with illness: forgone wage income (due to short-term disability). In this paper, we quantify the economic burden of illness in India attributable – separately – to wage loss and to medical care spending, as well as differences in them across the socio-economic distribution. Using data from two Indian longitudinal household surveys, we find that wage loss accounts for more than 80% of the total economic burden of illness among the poorest households, but only about 20% of the economic burden of illness among the most affluent. Overall, we find that wage loss accounts for a substantial share of the total economic burden of illness in India – and disproportionately so among the poorest households.

Works in Progress

Effect of Digital Jobs on Women’s Mental Health, Agency, and Social Norms

We run a randomized control trial to analyze the impact of digital job offers on women's mental health and empowerment in 2,700 married households in Mumbai, with a control group of 500 households. The study assessed the intent-to-treat effects on mental health (PHQ-9), an agency index, and a social norms index, while also considering the influence of job contract type. Results indicated no significant changes in systemic mental health or on agency and social norms for home-based jobs. However,  women working outside the home reported higher gains in agency and social norm measures, while those assigned home-based jobs experienced negative shifts in family norms concerning women working outside. Further, we find increases in self-reported tension related to household finances. Preliminary findings suggest minimal perceptual changes in husbands regarding their wives' work and agency, attributed to limited daily observation. 

LLM-ChatBots and Women’s Agency around their Sexual and Reproductive Health (with Azra Ismail)

[Status: Design stage]

This study examines the role of anonymous, empathetic technology interventions, such as those involving GPT-4, in shifting women's health-seeking behavior and its subsequent effects on their agency outcomes. Through a Randomized Controlled Trial,  we analyze if the anonymity of an LLM-bot, compared to other modalities, significantly increases women's willingness to discuss sexual and reproductive health (SRH) concerns, within the context of stigma, discrimination, and the quality of available services. The outcomes will provide insight into the value women place on privacy in SRH inquiries and the most conducive environments for enhancing agency and progressive attitudes towards women's SRH.

Increasing Gender Representation in AI: Harnessing Social Dynamics to Boost Women's Participation in India (with Akhila Kovvuri)

[Status: Design stage]

In the field of AI, gender diversity and representativeness in training data are critical for creating fair and unbiased models. We complement the government's efforts to diversify AI with a large-scale experiment on recruitment strategies to increase take-up by a diverse range of women across socio-economic class, age, education, and occupations. Can we increase the take-up of AI jobs among women in India by focusing on role-model and group-based recruiting strategies? To answer this research question, we randomly assign villages to either receive an individual or a group-based job offer (multiple women from the same village receive a job offer). We then observe the AI job take-up of other women in a subsequent round to understand whether offering more women jobs in the first phase affected take-up in the second round. 

Others

Developing an Agency Measurement Tool for Urban Slum Women 

[Status: Design stage] 

We propose a  new tool to capture women's agency in urban slums among women traditionally considered to be more empowered. We aim to capture the impact of social norms, labor market decisions, and health decisions on women’s agency, specifically in an urban slum setting. In this study, we investigate methods that effectively measure women’s agency, including self-report surveys, vignettes, games, implicit association tests, list randomization, network effects and digital patterns in interactions with digital healthcare tools to develop metrics with a limited set of questions on select parameters.