UCLA

This thesis studies a network formation model with incomplete information, which introduces the neighborhood effect into the analysis of network formation. We show that the model is identified under some mild conditions. To overcome the computational burden, we propose to use the nested pseudo-likelihood algorithm to estimate the parameters of interest. Finite sample performance of the NPL estimation method is investigated through several Monte Carlo experiments. We find that a positive neighborhood effect makes agents more likely to form links, which can increase the network density. Besides, we also discuss three potential research directions.