COVID-19, Income Shocks, and Women’s Employment in India

Abstract

ABSTRACTExisting evidence shows that the COVID-19 pandemic led to larger employment losses for working women in India. This article examines the heterogeneity that underlies these trends by studying the impact of income shocks due to the COVID-19 induced national lockdown (April–May 2020) on women’s employment. Using individual-level panel data and a difference-in-differences strategy that exploits the imposition of the lockdown and accounts for seasonal employment trends, the study finds that women in households facing a hundred percent reduction in men’s income during the lockdown were 1.57 pp (27 percent) more likely to take up work after restrictions eased (June–August 2020). These results are predominant in poorer and less educated households. However, these positive employment trends are largely transitory as the effect on women’s employment reduces to 13 percent in these households during September–December 2020. These findings underscore the use of women’s labor as insurance during low-income periods by poorer households.HIGHLIGHTSWomen’s labor acts as insurance during periods of men’s income loss.The increase in labor market participation is only observed for married women.Rural women participate in less-secure casual agricultural labor.Urban women access more secure fixed-wage work and self-employment.Increase in women’s labor force participation is mostly transitory.KEYWORDS: EmploymentCOVID-19income shocksgenderIndiaJEL Codes: J22J23J16 ACKNOWLEDGMENTSWe thank the anonymous reviewers for extensive comments and suggestions.SUPPLEMENTAL DATASupplemental data for this article can be accessed online at https://doi.org/10.1080/13545701.2023.2250797.Notes1 The national lockdown was imposed only during March 24, 2020–May 2020. Thereafter, only local lockdowns were imposed based on COVID-19 cases in a state or district.2 Marianne, Bertrand, Kaushik Krishnan, and Heather Schofield (Citation2021) and Marianne Bertrand et al. (Citation2020) discuss how some of the key indicators like employment, income, and consumption changed over time and across different categories of employment – self-employed, casual labor, fixed wage work – due to the lockdown in India.3 Empirical studies from developed countries show that women’s labor supply is pro-cyclical in aggregate (Joshi, Layard, and Owen [Citation1985] for the UK, Killingsworth and Heckman [Citation1986] for the US, and Darby, Hart, and Vecchi [Citation2001] for other OECD countries).4 See Sonia Bhalotra and Marcela Umana-Aponte (Citation2010) for evidence on a number of developing economies including India, Emmanuel Skoufias and Susan W. Parker (Citation2006) for Latin America, Elizabeth Frankenberg, James P. Smith, and Duncan Thomas (Citation2003) for Indonesia, Carola Pessino and Indermit S. Gill (Citation1997) for Argentina, and Joseph Y. Lim (Citation2000) for the Philippines.5 The broad strata are the homogeneous regions which are a collection of neighboring districts within a state that have similar agro-climatic conditions. Each homogeneous region (HR) is then divided into rural and urban sub-strata. The urban regions of an HR are further stratified into four strata based on town size. Thus, each HR has five sub-strata. From each sub-strata PSUs are selected randomly. Additionally, CPHS provides household-level sampling weights. We do not use weights in our analyses since there was attrition in the sampled households due to the pandemic. Our results, however, remain robust to conducting a weighted analysis. The results are available on request. For more details on the sampling strategy and the sampling weights, refer to CMIE’s documentation here.6 The excluded states are mostly inaccessible or difficult to access regions. These include the four border states in the Northeast – Arunachal Pradesh, Nagaland, Manipur, and Mizoram and some islands. Despite these exclusions, the survey represents almost 98.5 percent of the total population in India.7 The survey does not collect data on days and hours worked in 2019 and hence we cannot use the intensive margin of work as an outcome variable in our analyses. Also, in general, employment rates obtained using the CPHS data have been shown to approximate employment rates for women in the daily status of nationally representative data like national Sample Surveys (Afridi, Mahajan, and Sangwan Citation2021). Recently, CPHS was criticized for its systematic sampling strategy that over-samples well-to-do households. However, given that we are interested in heterogeneity across households and not aggregate trends, we believe this is not a major cause of concern in our analyses.8 It is possible that in households where men could not find employment even when the restrictions were lifted, women entered in sectors which were less affected. For example, home food delivery was common self-employment activity by women during the unlock months (Nagpaul Citation2020).9 The asset index is created using Principal Component Analysis (PCA) for multiple binary indicators depicting ownership of various assets including mobile, health insurance, LIC, bank account, PF account, Kisan credit card, credit card, and Demat account.Additional informationNotes on contributorsIshaan BansalIshaan Bansal is currently a graduate student studying MPA/ID at Harvard Kennedy School. Previously, he worked at IDinsight as a Senior Associate at IDinsight, a global nonprofit advisory, focused on helping global development leaders maximize their social impact.Kanika MahajanKanika Mahajan is Assistant Professor of Economics at Ashoka University, India. Previously, she has taught at the School of Liberal Studies, Ambedkar University, Delhi. She obtained her PhD in Economics from the Indian Statistical Institute in 2015, and her primary research interests include empirical development economics in the field of gender, labor, and agriculture.