一种估计连续策略博弈的两阶段方法

 最近实证博弈研究的迅速发展为分析市场中的策略互动、进行政策分析与反事实实验提供了有效的工具。本文提出一种估计不完全信息连续策略博弈的两阶段方法，它可以处理私有信息的影响。第一阶段通过非参数分位数回归估计局中人的策略与期望支付函数；第二阶段利用贝叶斯——纳什均衡不等式构造模拟最小距离估计量，最终获得结构参数的估计。数值模拟显示本方法有良好的小样本表现。与现有文献的嵌套固定点方法相比，本方法不需计算均衡，极大地降低了计算量，并减轻了多重均衡的干扰。本方法既可以用于估计离散状态博弈，也适用于连续状态博弈。

A Two-step Estimation for Incomplete Game with Continuous Actions

Recent developments in empirical game research provide useful tools for analysis of strategic interactions in market, policy recommendations, and counterfactual experiments. This article proposes an estimator for incomplete game with continuous actions, which copes with the influence of private information. It consists of two steps. In the first step, the strategy function and expected payoff function are estimated by nonparametric quantile regression. In the second step, the structure parameter is estimated by simulated minimum distance method. Numerical simulations are used to demonstrate the finite sample properties of the two-step estimator. This algorithm does not need to compute the equilibrium, thus alleviating computing burden. It could be used in both discrete and continuous state games.