Simultaneous equations with binary outcomes and social interactions
提出一个离散选择联立方程社会互动模型，基于不完全信息博弈给出微观基础，讨论识别条件并设计两阶段估计方法，蒙特卡洛实验表明该方法在有限样本和大网络中表现良好。
This paper introduces a discrete-choice simultaneous-equation social interaction model. We provide a microfoundation for the econometric model by considering an incomplete information game where individuals interact in multiple activities through a network. We characterize the sufficient condition for the existence of a unique BNE of the game. We discuss the identification of the econometric model and propose a two-stage estimation procedure, where the reduced form parameters are estimated by the NPL algorithm in the first stage and the structural parameters are recovered from the estimated reduced form parameters by the AGLS estimator in the second stage. Monte Carlo experiments show that the proposed estimation procedure performs well in finite samples and remains computationally feasible when networks are large. We also provide an empirical example to illustrate the empirical relevance of the proposed model.