Collaborative Research: Empirical Analysis of Static and Dynamic Strategic Interactions

战略互动模式对产业组织经济学有着深远的影响。一个企业进入寡头垄断市场的行为通常取决于其他竞争企业的进入行为。许多企业进入模型都有博弈论基础。理解博弈论模型的经验内容，即竞争企业的战略选择是相互依赖的，对于制定反垄断和管制的经济政策具有重要意义。战略竞争模型的信息结构可以根据企业是否最大化静态或动态预期利润，以及每个竞争企业是否拥有其他市场参与者不能观察到的私人信息来区分。由NSF拨款支持的几个研究项目旨在研究战略互动的计量经济学模型，在这些模型中，相互竞争的公司在有限数量的相互排斥的行动之间做出选择。第一个项目研究了私人信息假设下博弈论战略竞争模型的静态版本的识别和估计。开发了一个算法来计算这些模型中的整个平衡点集。将这些估计和计算方法应用于股票分析师推荐市场，发现了同业影响和多重均衡的实质性影响的有力证据。第二个项目将静态模型扩展到动态环境，在动态环境中，竞争企业在马尔可夫完美均衡中重复互动。分析结果表明，这些模型的结构参数可以从企业进入行为的数据中识别出来。辨识分析自然得到了具有连续和离散状态变量的动态寡头模型的灵活的非参数估计器和实用的半参数估计器。目前正在收集一组独特的石油勘探数据，以证明这些模型的经验预测。第三个项目允许一个完整的信息环境，在这个环境中，所有的公司和代理人通常都可以看到对公司利润的潜在冲击。为了便于估计具有多策略均衡和混合策略均衡的模型，提出了一种可行的计算算法。将该算法应用于加州高速公路采购拍卖中的进入行为，可以回收投标公司的显着进入成本和进入行为的显着多重均衡决定因素。这项研究的结果除了在产业组织领域外，还具有潜在的重要经济应用。在这些研究项目中开发的计量经济学方法将有助于解决应用微观经济学中的问题，在这些问题中，代理人的效用是相互依赖的，代理人之间存在战略上的相互作用。例如，公共财政中的同行效应和城市经济中的集聚。

Models of strategic interactions have a profound effect on the economics of industrial organization. The entry behavior of a firm into an oligopoly market typical depends on the entry behavior of other competing firms. Many firm entry models have a game theoretic foundation. Understanding the empirical contents of game theoretic models, in which the strategic choices of competing firms are mutually interdependent, has important implication on the economic policies of antitrust and regulation. The information structure of strategic competition models can be differentiated based whether firms maximize static or dynamic expected profits, and whether each competing firm possesses private information that can not be observed by other market participants. Several research projects supported by this NSF grant aim to study econometric models of strategic interactions where competing firms choose between a finite number of mutually exclusive actions. The first project studies identification and estimation of static versions of game theoretic strategic competition models under the private information assumption. An algorithm is developed to compute the entire set of equilibria in these models. Application of these estimation and computation methods to the market of stock analyst recommendations discovers strong evidence of peer influence and substantial impact of multiple equilibria. The second project extends the static models to a dynamic setting where competing firms interact repeatedly in a Markov perfect equilibrium. The results of the analysis show that the structural parameters of these models can be identified from data on the entry behavior of firms. The identification analysis naturally leads to a flexible nonparametric estimator and a practical semi-parametric estimator for dynamic oligopolistic models with both continuous and discrete state variables. A unique data set on oil exploration is being collected to demonstrate the empirical predictions of these models. The third project allows for a complete information setting where latent shocks to firm profits are commonly observable to all the firms and agents. A feasible computation algorithm is developed to facilitate estimating these models with multiple and mixed strategy equilibria. Application of this algorithm to entry behavior in California highway procurement auctions recovers significant entry costs by bidding firms and significant multiple equilibria determinants of entry behavior.Broader impacts. The results of this research have potentially important economic applications beyond those in the field of industrial organization. The econometric methods developed in these research projects will be useful for problems in applied microeconomics where the utilities of agents are interdependent and agents interact strategically. Examples include peer effects in public finance and agglomeration in urban economics.