【作者】 傅星；

【导师】 林寅；

【作者基本信息】 首都经济贸易大学 ， 数量经济， 2005， 博士

【摘要】 经济学作为一门单独的科学自从出现以来,已走过了数百年的发展道路,逐步形成了以概念、假设、数据以及数学模型为基础的比较完整的理论体系,并在经济学研究中占据了主流地位。然而,随着知识经济时代的到来,传统经济学遇到了空前的挑战。首先,由于现实经济中涉及的因素众多,相互关系复杂,给数学方法的运用带来了一定困难;其次,传统方法没有很好地解决经济系统的层次性结构问题,因此,无法对不同层次的特性差异给予一个整体性的解释,也就产生了宏观、微观相互脱节的理论体系上的缺陷;再次,传统经济理论对现实经济普遍存在的信息不完备与不确定性尚不能给予系统的解释;最后,传统经济中以观测数据为基础的经验主义的研究方法,无法深入了解经济的微观运行过程,从而降低了理论对现实的解释与预测能力。 从上世纪八十年代开始,世界各国的许多科学家对经济学作为一个演化的复杂系统展开了大量的研究工作,并在理论与方法上取得了一系列令人瞩目的研究成果。其中,基于复杂适应系统(CAS)理论的多主体经济仿真是经济系统复杂性研究中的一个重要分支。上世纪九十年代中期,美国Sandia国家实验室开发了基于CAS理论的经济模型ASPEN,该模型从仿真经济中微观个体的属性与行为出发,通过建立微观个体间的相互联系,间接地构造了美国经济的宏观构架,从而较好地解决了宏观与微观的联系问题。在对该模型的运行实验中,较好地再现了美国经济的波动现象,所进行的货币政策实验也取得了与经济理论相一致的结果。总之,该模型具有传统经济计量模型与投入产出模型所不具备的优势。 以ASPEN为代表的基于CAS理论的多主体经济仿真对我国的经济学研究同样具有重要的理论意义与应用价值。但由于美国未公布该模型的技术细节,国内一些单位,如中国人民大学信息学院、吉林大学数量经济研究中心等都针对该模型进行了大量的研究、开发工作,并取得了一定成果。 针对国内缺少ASPEN所需的大规模并行计算机的问题,作者开发了能够在微机局域网环境下使主体仿真程序并行工作的分布式多主体仿真平台DABS,并在该环境下成功地仿制了美国ASPEN模型。从目前所获得的国内研究资料分析,该模型在主体类型、主体数量、主体学习算法、主体问的相互联系以及模型的运行结果等方面,更加接近ASPEN原模型,在国内具有领先水平。在成功研制仿ASPEN模型的基础上,作者又开发了中国转型经济模型CASPEN。该模型重点仿真了计划经济与市场经济两种体制并存的中国转型期特有的经济特点,并对该市场环境下财政政策与货币政策的效果进行了对比实验。根据对实验结果的分析,提出了完善我国社会主义市场经济的一些政策建议。 本论文在复杂适应系统理论的框架下运用多主体仿真技术研究现实经济问题,其意义在于:首先,本课题是对经济学研究方法的一种探索。通过多主体仿真研究经济问题与传统经济学研究中广泛采用的演绎、归纳方法有所不同,它以复杂适应系统理论为依据,以多主体建模及模型实验为手段,通过对实验结果的分析,揭示经济系统的内在规律。这种方法与2002年诺贝尔经济学奖获得者Vernon Smith的实验经济学方法有共同之处。其次,在经济全球化与信息经济的时代背景下,传统经济学从理论到方法都遭遇到了空前的挑战。由于我国目前处于经济转型阶段,机制与环境随时都在发生着较大变化,传统经济学方法运用时会更多还原

【Abstract】 As a branch of science, economics has gone through hundreds years and formed an overall theoretic structure on the basis of conceptions, hypotheses, data, and mathematic models. Therefore it has been the mainstream in the fields of economics researches. However, the traditional economics is confronted with a tremendous challenge with the advent of the knowledge economics era. Firstly, many factors involved in economic practices and their complex relationships make difficult to the application of mathematic methods. Secondly, there are no reasonable solutions with the traditional methods to hierarchical structure problems of the economic system, and so all-around interpretation cannot be given to the characteristic difference between hierarchies, as a result, disconnections between macro theoretic structures and micro theoretic structures come into being. Thirdly, incompleteness and uncertainty of information are ubiquitous in the world, and they cannot be explained systematically by traditional economical theories. Lastly, microcosmic process of economy cannot be found out thoroughly with experimentalism methods of traditional economics based on observational data, thereby capabilities of explaining and forecasting reality with theories have been reduced.Since 1980’s, scientists in many countries have developed lots of researches on considering economics an evolutional complex system, and have born abundant fruits. Hereinto, the study of Multi-Agent simulation of economies based on the complex adaptive system theory is an important research branch of the complexity of economy. In the middle of 1990’s, an economicsmodel based on the complex adaptive system theory-ASPEN was developed by SandiaNational laboratory in America. The model built the macro structure of American economy though simulating properties and actions of micro agents and simulating interaction of these agents. Accordingly, the relationship between macro and micro is solved preferably. The fluctuating phenomena of American economy can be imitated properly by experiments with the model. Anyway, the model has many advantages, which are absent in econometric models and input-output models of traditional economics.The Technique of Multi-Agent economic simulation based on the CAS theory is equally of significant theoretic implications and application value in China’s economic studies. Because the detailed technique of ASPEN was not released publicly, some universities and institutes, such as the Information School of Renmin University of China, the Research Center of Quantitative Economics of Jilin University, have done lots of work involving the study of such model. Quite some progress has been made.Because of lack of parallel-computer needed by ASPEN on a large scale in our country, a distributed platform, named DABS, for Multi-Agent simulation was developed with microcomputer under the circumstance of local area network by the author. Based on this platform, an imitation of ASPEN was implemented, and the simulating results are much closer to the prototype of ASPEN considering agent types, quantities, learning algorithms and relationships among agents.Furthermore, a model of China’s transitional economy named CASPEN is developed successfully. Special characteristics of the transitional economy, with the planned economy and market economy concurrently existed, are simulated with the model, and then through comparing finance policy with monetary policy of market circumstance by experiments. By analyzing the experimental results, some suggestions about policy are presented to improve socialistic market economy of China.In this paper, the significance of research on some practical economic problems with Multi-Agent simulation based on the complex adaptive system theory is presented. First of all, it is an exploring of the methodology of economic studies. That is to say, study of Multi-Agent simulation of economies based on the complex adaptive system theory can uncover some intrinsic rules in economic system by analyzing experimental results. It displays the co更多还原