Panel Data for the Study of Network Economics and Risk Sharing

How individuals cope with adversity or find jobs, among other things, depends on the strength of their relationships within communities and across geographic areas. Research in many disciplines such as economics, sociology, computer science, and statistics therefore rely heavily on social network data. Collecting detailed network data on populations is very costly and, consequently, research often includes only a small part of populations in their analysis. This research project will develop innovative methods to inexpensively collect network data, use the method to create two large datasets on the socio-economic relationships among vulnerable populations. The researchers will match these network data with information on economic and labor market outcomes as well as physical and mental health status. In addition, the researchers will build statistical tools to facilitate the use of these data sets and make these datasets and the toolkit freely available to other researchers. The results of this research project will improve research on several topics such as social learning, risk sharing, and therefore improve policy making. The results will also increase the effectiveness of government and business policies.

This project will develop new methods to create network data, use the methods to build two large data sets on vulnerable people, and merge these data sets with data on several economic and social outcomes and make the data sets available to researchers. To study a large sample of vulnerable populations, the network relationships among them both within the local community and across large geographic regions, one needs a tool to make the data collection scalable. This project will use Aggregate Relational Data (ARD) to make this feasible. By asking individuals the number of people with a particular trait they are linked to (for various sets of traits), the PIs can estimate a network formation model that gives a picture of how community-to-community relationships vary across regions. ARD is very cheap to collect, in contrast with detailed, complete network data. Equipped with ARD and appropriate statistical and econometric techniques, researchers can sample numerous more respondents and create more network data. This research product will benefit theorists, econometricians, statisticians, and sociologists. The results of this research will allow researchers to study topics such as diffusion, social learning, multiplexing in networks, risk sharing, and social isolation, among others. It will allow researchers and policymakers to access a granular dataset that can better inform policies generally and particularly on vulnerable populations. The results are likely to improve the quality and targeting of policies.