离散型序贯博弈计量模型研究

发布时间：2018-01-20 08:36

本文关键词： 博弈计量模型离散型博弈序贯博弈完全信息完美信息　出处：《首都经济贸易大学》2014年博士论文　论文类型：学位论文

【摘要】：博弈论研究利益相互影响的决策问题，拓宽了经济学的研究领域，加深了经济学的分析。计量经济学揭示了经济活动中客观存在的数量关系，已在经济学中占据极其重要的地位。博弈论和计量经济学已成为当今分析经济问题的两种最有力的方法，都多次获得诺贝尔经济学奖。博弈计量模型是博弈论与计量经济学的结合，是对博弈模型的计量经济分析，融合了博弈论和计量经济学的优点，必将使问题的分析以更加符合现实的方式揭示经济活动的内在规律。博弈计量模型在经济研究中的作用越来越大，也引起了越来越多的关注。有不少学者在这一领域做出了贡献，但现有对博弈计量模型的研究，涉及序贯博弈的还不多。本文将对各种信息情况下离散型序贯博弈计量模型的估计进行系统讨论，包括离散型完全且完美信息序贯博弈计量模型的估计、离散型完全不完美信息序贯博弈计量模型的估计和离散型完美不完全信息序贯博弈计量模型的估计。对于每一类模型的估计，本文都是先从简单模型开始讨论，进一步引申到一般模型，最后给出一个实例，并在给定具体模型假设的基础上，求解所有均衡出现的概率，通过观测到的数据，估计模型参数。由于在博弈计量模型的估计中，研究者只能观测到行动，而不能观测到策略，本文在各种信息情况下离散型序贯博弈计量模型的估计中，都是基于行动而非策略分析博弈均衡，给出均衡行动组合满足的条件，并在保证博弈均衡唯一性的基础上，求解与博弈均衡一一对应的随机扰动项的范围。然后，通过计算随机扰动项在某范围内的积分和此范围内某均衡行动组合出现概率的乘积，就可以得到对应的均衡行动组合出现的最终概率，这个概率是用随机扰动项的多重积分来表示的。当博弈的参与人数量较少时，，这个多重积分的维数较低，一般可以直接计算，模型参数的估计也可以直接用最大似然法。但当博弈的参与人数量较多时，这个多重积分的维数较高，直接求积分的计算量会很大，本文用GHK（Geweke-Hajivassiliou-Keane）模拟的方法来简化计算，并用最大模拟似然法估计模型参数。博弈计量模型的研究有着重要的理论意义和现实意义，而序贯博弈又是现实中普遍存在的博弈类型，本文给出的离散型序贯博弈计量模型的估计方法，完善了博弈计量模型的理论体系，拓展了其应用范围。
[Abstract]:Game theory studies the decision-making problem of the mutual influence of interests, broadens the research field of economics and deepens the analysis of economics. Econometrics reveals the objective quantitative relationship in economic activities. Game theory and econometrics have become the two most powerful methods to analyze economic problems. Game econometrics model is the combination of game theory and econometrics, it is the econometric analysis of game model, which combines the advantages of game theory and econometrics. The analysis of the problem will reveal the inner law of economic activity in a more realistic way. Game econometric model plays a more and more important role in economic research and has attracted more and more attention. Many scholars have made contributions in this field, but the existing research on game econometric model. In this paper, the estimation of discrete sequential game econometric model is discussed systematically, including the estimation of discrete complete and perfect information sequential game econometric model. The estimation of discrete complete imperfect information sequential game econometric model and discrete perfect incomplete information sequential game econometric model. For each kind of model, this paper first discusses the simple model. Finally, an example is given, and on the basis of the hypothesis of the given model, all the probabilities of equilibrium are solved, and the observed data are obtained. Estimation of model parameters. Because in the estimation of game econometric model, researchers can only observe the action, but not the strategy, this paper estimates the discrete sequential game econometric model under various information conditions. All of them are based on action rather than strategy to analyze game equilibrium, and give the conditions that equilibrium action combination meets, and on the basis of ensuring the uniqueness of game equilibrium. The range of random perturbation terms corresponding to game equilibrium is solved. Then, the product of probability of occurrence of a combination of equilibrium actions is calculated by calculating the integral of random perturbation term in a certain range. The final probability of the corresponding equilibrium action combination can be obtained, which is expressed by the multiple integrals of the random perturbation terms. When the number of players in the game is small, the dimension of the multiple integrals is lower. Generally, the model parameters can be calculated directly, and the model parameters can be estimated directly by the maximum likelihood method. However, when the number of players in the game is more, the dimension of the multiple integral is higher, and the calculation of the direct integration will be very large. In this paper, the GHKK Geweke-Hajivassiliou-Keane method is used to simplify the calculation, and the maximum simulated likelihood method is used to estimate the model parameters. The study of game econometric model has important theoretical and practical significance, and sequential game is a common game type in reality. This paper gives the estimation method of discrete sequential game econometric model. The theoretical system of game econometric model is improved and its application scope is expanded.
【学位授予单位】：首都经济贸易大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：F224.32

【引证文献】