It is probably not an understatement to say that the policy response to COVID-19 in terms of social protection programs has been staggering. Colleagues have impressively been tracking this through the global monitoring platform. New programs are being created and existing programs are being expanded, and we continue to learn about how and for whom these programs can be effective to offset the negative impact of crisis. In their paper studying the world’s largest employment guarantee scheme in India, Afridi, Mahajan, and Sangwan examine the extent to which an existing workfare program cushioned the impact of the crisis. This program, the Mahatma Gandhi National Rural Employment Guarantee Act (MG-NREGA) program, mandates the offer of 100 days of manual work on public rural infrastructure projects (such as irrigation canals and roads). It also mandates a reservation of 1/3rd of jobs in each MG-NREGA project for women. In 2018, 76 million people worked in MG-REGA. And relevant for the estimations in Afridi et al, the scheme was not working equally well in every district (or state), as well documented in pre-crisis studies. One punch line of this paper is that how the program was performing before the crisis matters for the extent to which the program can offset job losses – not shocking but worth documenting.
They use panel household data (Consumer Pyramids Household Survey, CPHS) from 2019 and 2020. As an aside, there is not much information in the paper on the survey collection in 2020, though there is some focus on attrition in the paper. In particular, I presume the August 2020 data were collected with face-to-face surveys --a surprise since I have presumed no one was doing face-to-face surveys in August 2020. Using these panel data, they show clearly the V-shaped employment pattern from January 2020-August 2020, though with only a partial recovery after the lockdown such that levels in August 2020 remain lower than to pre-pandemic levels.
To examine the extent to which MG-NREGA offsets the negative employment effects of the crisis, they do not use contemporaneous workdays generated under MG-NREGA in 2020. We expect this to be endogenous to the scope of the crisis in the district. Instead, they substitute the 2020 levels of the program with coverage (days worked in the program per rural population) from 2014-2018. They label this historical state capacity. They apply a difference-in-different estimation where the first difference is employment Jan-Mar 2020 to April-Aug 2020, and the second is from the same periods in 2019. They also control for occupation-specific time fixed-effects to further control for job losses associated with employment structure before the crisis.
Before focusing on the MG-NREGA results, the paper finds that the employment impacts of the crisis are larger for men than women, though this gap falls as male employment recovers as mobility restrictions eased. But, as the authors mention, among those working before the lockdown, the loss of work is greater for women than men (as others have shown). This emphasizes the important distinction between measuring the rate of work stoppage (for those working before the crisis) versus a comparison of employment rates overall with pre-crisis levels. Still, these results seem unique from others I have seen that largely show that the employment declines are, when different, larger for women than men. Recall that India stands out in terms of gender gaps on employment and very low levels of female labor force participation (see Florence’s blog from January 2020 on 28%).
The main focus of the paper is whether districts with higher pre-crisis capacity in delivering workfare had lower employment losses than other districts. During the stringent shutdown (April-May), there is a positive impact on employment from historical capacity (“NREGA” in Table 2 below) compared to employment for the same period in the previous year. Combined with the estimates of overall rural employment losses, this suggests that employment losses in areas with higher MG-NREGA state capacity were substantially lower. This effect is larger and significant for the latter period, with less restrictions. There is, as we might expect, no effect of district MG-NREGA capacity on employment rates for urban areas.
Here’s the twist: The employment effect of MG-NREGA capacity in rural areas is only observed for female employment. [Although they do not find employment effects for men on average, they do find that higher capacity states offset job losses for less education men and men from poor households.] Given the size of the overall employment decline for rural women (1 percent point), the size of the impact suggests that women who were not working before the crisis then entered into work in historically high MG-NREGA state capacity areas. The authors note that this is consistent with the literature on counter cyclicality of women's labor force participation, but they also emphasize that the suitability of employment opportunities likely plays a role. The first question to come to mind is likely whether this because of the 1/3 reservation for women. The authors say no – the reservation does not seem binding since women workers made up for about 49% of work days, before and after the pandemic. The authors look for supply-side factors – specifically those associated with more restrictive mobility of women and, therefore, a higher demand for work in the village of the type offered by MG-NREGA. They examine being ever married or having small children. Indeed, these traits are associated with an even larger up-take in employment for women.
This paper touches on a number of important issues, among them importance of looking at subnational variation in program delivery. And, like most papers, if left me asking questions. Among these: why does capacity to deliver (as measured by the pre-crisis data on workfare) matters for women but not for men and to what extent program design explains the differential results among women with regards to marriage and presence of children?