The formation of networks with local spillovers and limited observability
研究了在信息溢出和有限观察半径下，个体如何形成网络，发现噪声大时小观察半径产生更大方差和更高总收益，并用物理学和经济学中的合作网络数据验证了模型。
This paper analyzes the formation of networks in which each agent is assumed to possess some information of value to the other agents in the network. Agents derive payoff from having access to the information of others through communication or spillovers via the links between them. Linking decisions are based on network-dependent marginal payoff and a network independent noise capturing exogenous idiosyncratic effects. Moreover, agents have a limited observation radius when deciding to whom to form a link. I find that for small noise the observation radius does not matter and strongly centralized networks emerge. However, for large noise, a smaller observation radius generates networks with a larger degree variance. These networks can also be shown to have larger aggregate payoff. I then estimate the model using a network of coinventors and scientific collaborations in physics and economics, and find that the model can closely reproduce a variety of observed patterns. I show that local search is important in all the empirical networks considered, but that economists tend to search more broadly for new collaboration opportunities.