A1046
Title: Identification of social effects through variations in network structures
Authors: Ryota Ishikawa - Waseda University (Japan) [presenting]
Abstract: Identification in linear social interaction models is studied through network structures. Existing identification results focus on cases where many identical networks are simultaneously observed within the same dataset. In reality, many networks with different structures are repeatedly observed within the same dataset. Conventional identification strategies address such heterogeneous networks by constructing a single large network through the stacking of different network structures. This approach implicitly assumes that a single large network is repeatedly observed, which is not consistent with network observations. Identification conditions are developed in settings where multiple networks with different structures are simultaneously observed within the same dataset. The proposed conditions are consistent with network observations. If the identification conditions are satisfied, the minimum network size is smaller when observing multiple distinct networks than when observing many identical networks. Examples of networks satisfying the identification results are illustrated.
