India has a particularly low level of female labor force participation, clocking in at 28 percent in 2015. And this was down from 37 percent in 1990. An interesting new by Erica Field, Rohini Pande, Natalia Rigol, Simone Schaner and Charity Troyer Moore looks at how women’s control over earnings can increase their participation in the labor market (and move social norms) in the context of one of the largest public works programs in the world: the Mahatma Gandhi National Rural Employment Guarantee Scheme (MGNREGS).
Field and co. are working with communities in northern Madhya Pradesh – an area with low female labor force participation. The way MGNREGS works (simply put) is that folks are guaranteed a certain amount of public works work for a set wage. The innovation here comes in how they are paid. In 2012, the Indian government decided that this should go directly into an individual’s bank account. But for a time, people, and especially women, didn’t have these bank accounts. So, Field and co. start with a basic (randomized) treatment of helping women set up a bank account. In another group, they add to this direct deposit of the MGNREGS payment. Finally, they cross-randomize training on how to use bank accounts. So far, so interesting.
But the real world will always throw twists into even the best planned policy experiment. Field and co. faced a doozy: in late 2017, the Indian government made a massive push to get everyone a bank account. And indeed, 63 percent of the control group (which was unbanked) gets an account. So, Field and co. are going to focus only on the effects of getting direct deposit and training versus just having an account. Think financial control, not financial inclusion.
In terms of data, Field and co. have a baseline census (with minimal information) and 2 rounds of a follow up survey conducted one year and three years after the first of accounts were opened. They then bring in administrative data on bank accounts and from MGNREGS.
Before getting to results, Field and co. introduce a model of intrahousehold allocation. I am not going to reproduce it here, but the key point that makes it interesting is that they introduce the idea that violating norms will cost people (in terms of utility). And these costs come from 2 places: going against personal beliefs about what a woman should do held by the woman and/or her spouse and costs incurred by going against the prevailing belief in the community on whether a woman ought to work or not. These social norms figured prominently in a recent paper on Saudi Arabia and women’s labor force participation that I about back in May. And as in that paper, both social norms and personal beliefs can change when people can update their information.
So, what do they find? First of all, Field and co. don’t have power to look at the effects of the direct deposit vs the training, so we are going to focus on the impact of direct deposit and training. Also, they expect results to differ by women’s prior labor market engagement – so all of the results look at heterogeneity by whether the women had worked for MGNREGS in the past or not. It turns out this is correlated with having other kind of work experience as well as earnings so let’s call these experienced vs. not workers (which is a bit too extreme, but you get the idea).
Field and co. start by looking at how folks interact with the financial sector. And they find that the treatment (direct deposit plus training vs. just an account) leads to more autonomy in engaging with the financial sector as measured by an index of different indicators (e.g. going to the bank alone and being comfortable with this). This effect is more pronounced for women without MGNREGS work experience.
The treated women also work more. Looking at the whole sample, there is a significant increase in an aggregate labor supply index in the short run (1 year) which is no longer significant in the longer run (3 years). But when they focus on the not experienced women, they can see not only a strong short run effect, but also a significant one in the long run as well. To make things more interesting, the effect is particularly strong for the private labor supply sub-index, this stays significant in the long run, while the MGNREGS index loses significance over time. To sum it up: the treatment helped the women with less experience at baseline move not only into the public works program, it then helped them stay in private sector work (there is a cap on the days in the public works, so doing both is definitely possible).
Does this increased participation in the labor market move the needle on measures of women’s empowerment? Here Field and co. focus on 3 types of indicators: her ability to buy things with her own money, her mobility, and her control over decision making over her own income and benefits payments. For the whole sample there is nothing much going on. For the not experienced women however, there is significant movement in the overall index in both the short and long run. Breaking this down, most of the action appears to come from purchases from her own money, although there are is a significant increase in mobility in the long run. Interestingly (and with some implications for how we might think about intrahousehold allocation) there is no significant movement in decision making.
On to norms. Field and co. do a lot of work to measure norms, including rolling out vignettes. They talk about norms in a way that’s different from the way I was taught to think about them, so I’ll try and use their terminology. First is “actual norms” which comes from personal preferences as well as vignettes on whether a working woman is a better wife, mother, caretaker. They roll these into an index and see a significant increase for treated women, but not for their husbands. The effect is concentrated among the not experienced women. The second group is “perceived norms” where Field and co. include the fraction of community who judge women who work (as reported by the respondent) as well as vignette data on community respect. Here there is no significant impact of the treatment for the whole sample of women, and a significant (at 10 percent) impact for the not experienced women. However, there is more action for the men: for the overall sample, there is a significant liberalization in these norms.
So this is quite an interesting set of results. And this paper sheds some light into how policy might be able to move the needle – both in terms of increasing labor force participation and shifting (some) norms for both women and men.