导　　师： 汤兵勇

学科专业： 081101

授予学位： 博士

作　　者： ;

机构地区： 东华大学

摘　　要： 上个世纪后半叶,现代金融学高速发展,以有效市场理论、投资组合理论、资本资产定价理论和期权定价理论为基石的现代金融学理论大厦迅速建立,并由此推动了80年代以后金融实践的巨大发展,金融理论几乎完成了它的使命。但是越来越多的异常/(anomalies/)现象,如:金融危机的频发、市场过大的波动、收益率分布的肥尾现象、规模效应、反应不足和反应过度等等与有效市场假说不符,引起了跨学科的研究者对金融理论的重新思考,产生了许多的新理论新思想,试图完善或重建金融理论体系。 新金融涉及多个领域,如心理学、物理学、复杂适应系统和人工智能等。代表理论有行为金融学、分形市场假说、金融物理学和在复杂适应系统概念下的多智能体计算经济学。对新金融理论研究之后,本文认为金融学新体系的建立,其核心的内容便是对基础的权益/(股票/)市场动力学机制的确立。市场动力学分析市场最基础的影响因素以及其运动规律和相互作用,如预期的特征、决策的规律以及价格形成机制等。 本论文的主要工作和结论如下: 1、论文在第一个层面上讲是对证券市场动力学机制的研究。对已有金融理论动力机制的分析和新动力机制的提出及论述贯穿整篇论文。在研究的基础上,论文把现代金融学和行为金融学所暗含的动力机制模式归类为静态机制,把引入复杂适应系统后建立的智能体模型所揭示的新机制归为动态机制。本文认为静态机制下的模型,特别是行为金融学模型和动态模型并不是对立的,前者是后者在特定市场状态下的静态描述或称定格分析。 现代金融学的动力机制简单描述为:建立在投资者同质预期的理性预期均衡。其特点是:理性且同质的投资者、演绎的逻辑、均衡的机制,并由理性套利保证了市场的有效运行:行为金融学并没有抛弃演绎逻辑的均衡机制,其动力机制的新特点是投资者的异质性、有限理性及心理对效用函数的异化;复杂适应系统的思想启发我们提出了新的动力学机制:多样化的投资者以归纳逻辑决策形成的动态均衡系统,这里投资者不是同质也不是简单异质,而是多样化。非完美的信息让投资者不得不用归纳逻辑决策,动态地进行学习和适应,市场成为一个动态进化的系统。本文认为这是一个更加完整和接近实际的动力学机制。 2、本文对市场动力学机制的分析着眼于对市场异常的解释。已证实存在的市场异常是对有效市场假说的最重要的挑战,解释异常是新金融各种思想的起点。本文的研究是在静态模型和动态模型的框架内进行的。 在静态模型部分本文详细讨论了行为金融学对异常的解释以后,提出自己的模型: 东华大学博士学位论文 投资者信念模型。该模型包括一个定性的模型和一个解析的静态模型,模型吸收了动态 模型的思想,如归纳推理和学习,但仍然建立在均衡的框架上。定性的模型可以分析价 格和交易量的形成机理及相互关系,是对市场动力机制的新理解;静态的解析模型分析 了在某种特定状态下价格是如何在事件信息的影响下出现反应不足和反应过度的,同时 也可以解释其它一些异常,并预言了两种有待验证的异常。 在动态模型部分,本文引入并改进少数派博弈模型,建立了基于智能体的人工市场 模型。对少数派博弈模型的改进使智能体有了更高的“智能”、更强的适应性、更接近 真实市场投资者的特征,使少数派博弈模型成为分析资本市场异常的工具。在此基础上 建立的模型仿真出了很接近真实市场的收益率分布的尖峰、肥尾现象。 3、本文对市场动力学机制的分析之后,提出了短期预测的思路,并进行了实证研 究。预测性一直是争论的焦点。传统的思想认为证券的价格受大量的几乎无法处理的外 生因素的影响,这就导致预测陷入泥潭,因为无法处理这些巨量数据。而实际上,所有 这些因素都被投资者所吸收而形成投资者信念,包括对收益的预期及对风险的评估,因 此价格其实只受一个因素的影响,那就是投资者信念。价格和交易量的波动中必然包含 投资者信念的信息,从系统的观点看,这些信念是在变化的,但是变化存在响应时间, 只要挖掘出信念信息,短期预测在理论上存在可能性。 在静态模型的体系中,本文用神经元网络挖掘投资者信念来尝试预测,实证显示存 在一定的预测能力。在动态模型中,智能体的策略集中包含了智能体的决策信息,本文 发现让智能体模型用实际数据进行训练可以产生稳健的预测能力。 4、本文还致力于建立更合理的智能体整体市场模型。复杂适应性系统思想下的基 于智能体的计算模型是一种研究证券市场的新方法,其代表便是圣塔菲研究所的人工市 场模型。本文从市场动力学的角度认为圣塔菲的模型有许多并不完善的地方,如智能体 智能不够高、学习的速度不够快、价格的决定机制依然采用预期均衡机制、仿真结果不 够理想等。本文以扩展的少数派博弈模型为核心,建立一种新的基于智能体的计算模型, 本文的模型有更接近实际的结构和机制,特别是可以分析趋势交易者的形成和影响,模 型仿真认为趋势交易者是因为归纳推理及适应而自然产生的,这有别于行为金融学的固 定非理性假设;智能体同时拥有价值策略和趋势策略会更“聪明”,可以比纯价值策略 的智能更多获利;在对称信息下,趋势交易的存在不会扩大波动,反而可以有减少波 In the later half of the last century, the modern finance including efficient market theory, portfolio management theory, capital asset pricing theory and option pricing theory was built rapidly and soon applied to many new practices of finance successfully. The modern finance was on the way to success. But in recent years a body of evidence on anomalies such as frequency of finance crises, excess volatility, fat tailed distributions of returns, size effect, under- and overreactions, has presented a sharp challenge to efficient market hypothesis /(EMH/). Anomalies are attracting significant interest across a broad range of disciplines such as psychology, physic, complex adaptive system and artificial intelligence. The representative theories of new finance are behavior finance, fractal market hypothesis /(FMH/), econophysics and agent-based computational economics /(ACE/).After studied on the literatures, we point out that the key issue of new finance theory is building the dynamics mechanism of security /(equity/) market. The dynamics mechanism analyzes the fundamental ways of actions and interactions of elementary factors, for example, the mode of investor's expect, the law of the decision and the mechanism of the price, etc.The main work and conclusion are as follows:1. Define and study the dynamics mechanism of security /(equity/) market. This work is the main thread of our paper. Analyzing and representing both the old mechanism implied by traditional theories and new mechanism we proposed get through the whole paper. We classify the old mechanism as static mechanism and the new mechanism based on complex adaptive system as dynamic mechanism. We also point out that models under the static mechanism especially behavior finance models are not contradictory to, but snapshot-analysis of those under dynamic mechanism.The dynamics of modern finance can be simply described as rational expectations equilibrium, which is characterized as homogeneous expectations, deductive reasoning, equilibrium mechanism and supported by rational arbitrage. The dynamics of behavior finance is characterized as heterogeneous investors' expectations, bounded rationality and psychological bias of normal utility function; but it does not abandon deductive reasoning and equilibrium mechanism frame. Complex adaptive system motivates us to present the new dynamics mechanism: an evolving equilibrium system based on diversified inductive reasoning. In the new dynamics investors are neither homogeneous nor heterogeneous but diversified. Because of the imperfect information, diversified investors have to make their decisions by inductive reasoning, and keep learning to adapt to the evolving market. This is a more unified and more realistic dynamics mechanism.2. Our study on dynamics of securities market focuses on explaining the anomalies because the existing evidence of anomalies is the main challenge to efficient market hypothesis. Our work can be divided into two parts, which we term "static model" and "dynamic model".In the part of static model, we propose an investors' beliefs model that also can be classified into behavior finance. The new model including a qualitative sub model and an analytical sub model is yet based on expectations equilibrium mechanism but absorbs ideas from complex adaptive system, for example, inductive reasoning and learning. The qualitative sub model can explain the dynamic interaction between price and trade volume, this is valuable to understand market mechanism. Analytical sub model can explain how a certain pattern, based on that investors form their beliefs, can causes both underreaction and overreaction in different phases when event come. Additionally, our analytical sub model can explain some other anomalies and predicts two new anomalies that need to be tested by empirical analysis.In the part of dynamic model, we introduce and extend minority game, and then build an agent-based artificial market. In the extended new model, agents are more intelligent, more adaptive and more similar to the investors in real market. So m

关 键 词： 证券市场动力机制 复杂适应系统 市场异常 智能体建模 少数派博弈

分 类 号： [F830.91]

领　　域： [经济管理]