Network Statistics and Modeling the World Trade Network: Exponential Random Graph Models and Latent Space Models.

摘要

Due to advancements in physics and computer science, networks have becoming increasingly applied to study a diverse set of interactions, including P2P, neural mapping, transportation, migration and global trade. Recent literature on the world trade network relies only on descriptive network statistics, and few attempts are made to statistically analyze the trade network using stochastic models. To fill this gap, I specify several models using international trade data and apply network statistics to determine the likelihood that a trade tie between two countries is established. I also use latent space models to test the 'geography is dead' thesis. There are two main findings of the paper. First, the "rich club phenomenon" identified in previous works using descriptive statistics no longer holds true when controlling for homophily and transitivity. Second, results from the latent space model refute the 'geography is dead' thesis.