There has been an exploding interest in the role that social networks play in economic interactions. Developing economies provide an excellent testbed to gain insights about how social networks facilitate economic (see, e.g., Kremer and Miguel (2007), Conley and Udry (2010), Banerjee et al. (2013), Ambrus et al. (2014) for applications to social learning as well as risk-sharing). However, empirical research has been hamstrung by the cost of network data collection. There is a natural tension here. Because networks-based research questions are naturally about interdependence, data from many independent village networks are required for valid statistical analysis. At the same time, a nuanced analysis of social networks requires getting a complete and global picture of the network structure; thus, a survey of the entire community is needed and this can be time-consuming. Researchers typically either have very incomplete network data across many networks or meticulous data across a handful of networks. We propose using a method that reduces the collection of network data tremendously (we estimate 1/6 the cost) so that complete data can be obtained across numerous communities with ease. We apply this to the study of risk-sharing networks. A major factor that limits the extent to which villagers form risk-sharing networks is the fact that parties may not be able to observe each others? incomes (Cole and Kocherlakota (2001)). With increased urbanization and the growth of temporary migration, villagers may be unable to see increasing shares of other households? incomes, which may erode the risk-sharing network. We propose conducting complete network surveys across 100 villages, and then offering an income monitoring service, varying whether or not we reveal income realizations to various community members. We will study (i) the extent to which individuals are better (or worse) able to cope with shocks when they opt for the monitoring service, (ii) how the risk-sharing network evolves in response to our experimental conditions, and (iii) which members of a risk-sharing network are most/least incentivized to migrate.
In this project, we seek to understand how the monitoring function of the social network interacts with the ability to share risk. By randomly revealing information about a temporary migrant?s income to members of the migrant?s home village risk-sharing network, we can cleanly to study how this affects the risk-sharing network. If the information friction is classical, a la Cole and Kocherlakota (2001), committing to reveal the income would increases the scope for insurance. However, if individuals face time-inconsistent temptations to honor requests for transfers from their social networks, predictions are less clear-cut (Banerjee and Mullainathan (2010)). Despite the importance of hidden income, we know very little about how information about incomes actually passes through village social networks. Do individuals feel compelled to hide income information? Does knowledge about income shocks unintentionally diffuse through the network? Thus, the first piece of data we will collect includes a detailed ethnographic survey that will help us understand the perceived incentives to hide/spread income information faced by villagers. The aforementioned network data will be collected using the following protocol. On the first day a team will collect a complete village census, as is typical in many field-based development projects. The key difference is that we will take pictures of the houses as well as any present household members during the census. For a minimal time-cost, we are able to create a facebook of the village. Thus, on the second day, our staff can ask standard networks survey questions used in the literature with the aid of this facebook, which allows us to bypass paper data-entry, minimizes name-matching errors, and saves an immense amount of time relative to using a standard codebook mechanism.
