Social Network Theory, Informal Insurance and the Diffusion of Microfinance

The access to informal insurance and credit markets is strongly determined by a village's social structure, and in particular by the network of relations among villagers. Tranfers in money and in kind usually follow social links, whether they are formed by kinship, gender, occupation or simple friendship. Information about agents' needs and capacities is more easily acquired through the village social network. In the reverse direction, information about new technologies or opportunities goes through word of mouth and the diffusion of new methods is clearly shaped by the village social network. While  the importance of social networks and social capital has long been recognized,  recent advances in social network theory have transformed our thinking about these issues, and provided new insights which ought to be incoporated in the design of social transfer and microfinance programs.

Consider the following design problem : in order to insure villagers against idiosyncratic risk (like illness or accident), an NGO devises a transfer scheme where villagers in need receive money from other villagers, in order to equalize as much as possible consumption inside the village. While the problem may sound combinatorially intractable, Bramoullé and Kranton have shown, in an article published in the Journal of Economic Behavior and Organization in 2007, that bilateral contracts are sufficient to ensure full consumption sharing in the village. By a succession of bilateral money exchanges, any tranfer scheme can be achieved, as long as the network is connected so that no agent is excluded from the village social network. Bramoullé and Kranton overlook an important aspect of informal insurance schemes : reciprocity. An agent can only be incentivized to make a transfer today if she believes that she will receive a transfer tomorrow if the need arises. Bloch, Genicot and Ray, in an article published in the Journal of Economic Theory in 2007 show that, when reciprocity is taken into account,  the archtecture of the social network matters in the success of informal insurance schemes. They show that the key structural element of the network is the length of cycles : if cycles are short, it is easier to transmit information about agents reneging on their promises, and insurance transfers are more likely to be sustained. In order to shorten cycles in the social network, new links have to be created between carefully chosen targets in the village. Ambrus Mobius and Szeidl, in a paper forthcoming in the American Economic Review, link the agents' incentives to transfer to what they term 'social colateral' : by reneging on his promise, an agent loses the benefits of interacting with the agent he cheated. With this notion of punishment, networks supporting informal insurance are characterized by their degree of expansiveness, a measure of the average number of connections between any group of agents in the network and the rest of the community. In order to increase transfers in the village, one needs to make connections more fluid, by guaranteeing that no agent concentrates intermediation power and adding connections to eliminate bridges in the social network.

When transfers are in kind, they typically take the form of favors which are exchanged in the village. A favor differs from a monetary transfer as it is pair-specific : if a villager needs to use a specific tool, there may be only one other villager who can lend it to him. The structure of social networks which support favor exchanges differs from the structure of social networks supporting money transfers. In a paper published in the American Economic Review in 2012, Jackson, Rodriguez-Barraquer and Tan show that a network supports favor exchange when every link is supported, meaning that any pair of agents who are linked have a common friend. This particular network structure – termed 'social quilts' by Jackson, Rodriguez-Barraquer and Tan, involves a hierarchical union of clusters, but does not require a high level of clustering. Data from Indian villages corroborate the theoretical findings, showing that village networks may have a low level of clustering, but typically involve a high fraction of supported links.

Networks can also be used as information channels to diffuse innovations. In a recent experiment conducted in Indian villages in Karnataka and reported in a paper by Banerjee, Chandrasekhar, Duflo and Jackson just published in Science, a lending group used the village social network to disseminate information about a microfinance program. Initially, the villages were surveyed to map the social links among agents, based on different dimensions like pairs of agents going to the temple together, or pairs of agents lending to one another. In  a second step, one individual of the village was targeted as the seed from which information about the progam would be diffused. One of the objectives of the experiment is to identify how the speed of diffusion of the information depends on the centrality of the agent chosen as seed. The experiment led to two striking conclusions. First, the 'correct' notion of centrality which explains best the diffusion of information is the Katz prestige centrality measure, where an agent's centrality depends on the centrality of his neighbors. Second, the speed of diffusion does not depend on the number of agents who actually participate in the program, but simply on the number of agents who are informed about the program. Even though incentives to transmit information differ between participants and nonparticipants, this difference has little global effect on the diffusion of microfinance. These results clearly show how sophisticated centrality measures borrowed from social network theory can help NGOs target specific individuals in villages in order to improve the efficiency of social programs in developing countries.
 

References

• Attila Ambrus & Markus Mobius & Adam Szeidl, 2007. "Consumption Risk-sharing in Social Networks," NBER Working Paper No. 15719, February 2010
• Abhijit Banerjee, Arun G. Chandrasekhar, Esther Duflo, and Matthew O. Jackson, The Diffusion of Microfinance,Science , 26 July 2013: 341 (6144), 1236498 [DOI:10.1126/science.1236498]
• Francis Bloch, Garance Genicot & Debraj Ray, 2008, "Informal insurance in social networks," Journal of Economic Theory, Elsevier, vol. 143(1), pages 36-58, November.
• Yann Bramoullé & Rachel Kranton, 2007. "Risk-sharing networks," Journal of Economic Behavior & Organization, Elsevier, vol. 64(3-4), pages 275-294.
• Matthew O. Jackson & Tomas Rodriguez-Barraquer & Xu Tan, 2012. "Social Capital and Social Quilts: Network Patterns of Favor Exchange," American Economic Review, American Economic Association, vol. 102(5), pages 1857-97, August.