Collaborative Research: New Techniques for High-Dimensional and Incomplete Network Data

Northwestern University, Evanston IL

Investigators

Abstract

This research project will develop new statistical and econometric resources for researchers working with social network data. Social networks often play a central role in shaping the outcomes and decisions of economic agents. For example, they can predict the diffusion of a new technology or spread of a contagion, determine the influence of key agents in crime or research production, or account for treatment spillovers in many social programs. However, incorporating social network data into empirical work can be difficult. Networks are fundamentally high-dimensional objects, and often it is unclear which features will be relevant for a particular social or economic phenomenon of interest. In addition, the researcher may only observe some of the relevant connections between agents in the social network for reasons of privacy or cost of data collection. This project will develop new techniques for incorporating social network structure into econometric and statistical modeling that accommodates these limitations. The project will train graduate students and develop a new course on the use of network data in causal inference. Software also will be developed and disseminated. This research project will build new econometric methodology for models with high-dimensional and incomplete social network data. First, the project will develop a new nonparametric regression framework for networks, characterize its statistical properties, demonstrate its use for empirical work in the social sciences, and provide software for implementation. Second, the project will investigate the informational content of partial network data generated by referral-based sampling and contingency-table data. The project will develop new theory, methodology, and empirical practices for identification, estimation, and inference using these nonstandard data types. New insights will be provided into exactly what features of social network structure are relevant for economic outcomes and decision making and when and how this information can be recovered from incomplete data in practice. The results of this research will help empiricists better understand the role that networks play in economics and the other social sciences. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.