Social networks matter. We know that they affect decisions on whether and where to migrate (McKenzie and Rapoport, 2010), whom to marry (Banerjee et al., 2009), and which technologies to use (Conley and Udry, 2010). Sometimes they even restrict the choices available to us (Skoufias, Lunde, and Patrinos, 2009). What we do not know enough about is how network effects work. As a result of this knowledge gap, we do not know how networks may interact with policy, such as whether networks change in the face of government interventions. And we do not know if networks can have an empowering (or disempowering) effect on its members – an effect that can be measured through a member’s relative influence on intrahousehold bargaining. Can networks determine who gets how much of the household’s pool of resources?
In India, the hierarchical structure imposed by the caste system means that social networks are often restricted by caste. The potent combination of a patriarchal society and the caste system can limit women’s interactions to a small subset of the community. Examples that challenge these strict social norms, such as access to outside role models, have been demonstrated to improve women’s bargaining power (Jensen and Oster, 2009).
In the north Indian state of Uttarakhand, a government intervention called Mahila Samakhya aims to empower women through education. This intervention provides formal, informal, and vocational education, and organizes support groups with the explicit aim of empowering women to have a greater say in their households and communities.
As part of my dissertation, I surveyed 487 women in Uttarakhand communities and collected data on women’s networks and on participation in Mahila Samakhya. Results suggest that not only does the program empower participants, but also has significant spillover effects on non-participants. These findings in turn led my co-author Kathy Baylis and I to study how the program effects interact with social networks. We find that the program significantly diversifies networks: participants tended to have more friends of other castes than women living in untreated districts.
I also collected data on a series of empowerment measures such as ability to leave the house without permission and participation in off-farm employment. While diverse networks may be desirable in and of themselves (Behrman, Kohler, and Watkins, 2002), we also wanted to know whether– and how– friends’ participation in this program might empower a woman to have greater decision making abilities with respect to household resources. Answering these questions is tricky because of two potential sources of endogeneity. First, we need to account for the fact that women choose whether to participate in Mahila Samakhya. We control for this source of endogeneity by using the number of years a woman has lived in a village with the program (zero for untreated villages). Second, we recognize that the formation of networks is also endogenous. We instrument for endogenous networks using the number of other women in the village with a similar time to collect water and the number of other women in the village of the same caste. After weighting our instrumental variables by the structure of networks, our results show that friends’ participation in Mahila Samakhya empowers women– controlling for their own participation status– to engage in outside employment opportunities, and to leave the house without permission.
Not only do individuals learn new information from their friends (social learning), but they also adopt certain behaviors because they are influenced by their friends, particularly in support group-like settings (social influence), or gain utility from behaving like their friends (identity utility) (Montgomery and Casterline, 1996). We use proxies identified in field tests to decompose the overall network effect into these component mechanisms.
We use low educational attainment as a proxy for social learning, since the women who are least exposed to information are most likely to benefit from social learning. We also argue that the influence of support groups most affects women with low bargaining power, and use the spousal age ratio as a proxy. Our proxy for identity utility is how much a woman cares about her social group's opinion of her; the proxy works through the potential utility gains from associating more closely with the group. And we account for the strength of the norms faced by a woman via whether the husband cares a lot about villagers' opinion of him.
We find that social learning and social influence increase participation in the Indian National Rural Employment Guarantee Scheme, as well as a woman’s ability to leave the house without permission. Finally, we look at how social networks might influence a woman’s parenting practices, and find that social learning can significantly improve the food intake of non-participants’ children.
Our results highlight the importance of social networks and suggest that female empowerment and child nutrition interventions may benefit from accounting for social learning and influence. For example, programs that rely on social learning could just target a small number of older, well-connected women in a village in order to disseminate information through their existing social networks. Such programs might include interventions to inform mothers on the importance of immunization, or meetings to discuss coping mechanisms for victims of domestic violence. On the other hand, programs that rely on social influence or identity utility, such as interventions aimed at women to be employed together in artisan cooperatives, may work best if they target clusters of villages to build groups that increase women’s utility from associating with the group and increase friends’ influence on each other.
Blogger Bio: Eeshani Kandpal is an Indian national who is currently a consultant in the Education Sector of the World Bank's Human Development Network. She has a PhD and MS in Agricultural and Applied Economics from the University of Illinois and a BA in Economics and Classics from Macalester College, Minnesota. Her research focuses on how government interventions and peer networks can affect early childhood development and female empowerment, particularly among the worst-off. In this research, she uses tools from the program evaluation literature and spatial econometrics.