【作者】 张征争；

【导师】 黄登仕；

【作者基本信息】 西南交通大学 ， 管理科学与工程， 2009， 博士

【摘要】 本文以经典委托代理理论和有关心理学理论为基础,以CEO的过度自信心理偏好为出发点,以委托代理关系中股东和CEO的决策行为为核心,以努力产出风险、努力可观测性、风险态度为主线,从理论上建立模型深入探讨纳入CEO的过度自信心理偏好后,股东对CEO的薪酬合同设计问题,并采用实验经济学方法,分别设计经济学实验,对考虑CEO过度自信的理论模型进行验证。过度自信是“心理学中最稳健的发现”,是影响经济人进行决策的主要情感因素和心理因素之一。由于在真实的市场环境中无法精确测度人的过度自信程度,虽然对过度自信进行研究的理论层出不穷,但如何利用实证数据检验这些理论的有效性仍然面临着较大的障碍。根据Berg等(1986)的思想,引入实验方法,通过控制交易单位和概率的关系,实验者可以通过控制任意不同形式的偏好来引导参与人的行动,从而使得参与人表现出他们确实具有这种偏好时的行动。因此,本文立足于拓展经典委托代理理论“理性人”的假定,建立以“过度自信”为前提假设的理论模型,在充分借鉴西方实验经济学研究成果的基础上,利用计算机实验操作系统,根据理论假设来控制实验条件,选取真实的参与人,在虚拟的委托代理情景中进行决策,并通过分析实验数据,对基于过度自信的委托代理理论进行检验。在影响委托代理关系的众多复杂因素中,本文选取了努力产出风险、努力可观测性、风险态度等关键因素作为研究对象,集中探讨当这些条件发生变化时,委托代理双方如何进行决策。在对每一个条件进行研究的过程中,本文遵循的思路是:首先建立不同条件下股东和过度自信CEO的委托代理模型,然后根据理论假设进行实验设计并操作实验,最后对实验数据进行分析得出结论。本文基于以上思路展开研究,并得出基本结论,包括:1.在多轮重复实验中,参与人的决策向预测的最优行动收敛,且理论模型预测的最优行动和非最优行动有显著差异,理论模型在实验中得到有效验证,具有一定的参考意义;2.过度自信和风险态度无显著相关性,男性比女性更偏好风险和过度自信:3.在努力产出风险、努力可观测性、CEO的风险态度这些条件变化时,股东对不同过度自信程度CEO的薪酬合同设计各不相同。也正是围绕上述结论,本文形成以下2个主要创新点:1.在理论上,在第3、4、5章中,拓展了经典委托代理理论的基本假定,即CEO是“理性人”的假定,将CEO的过度自信心理偏好纳入委托代理理论框架,并着重从努力产出风险、努力可观测性、风险态度三个方面,系统研究了不同过度自信程度CEO在不同条件下的最优薪酬合同设计,不仅丰富了经典委托代理理论,而且更清晰、客观地描述了股东和CEO的委托代理关系。2.在方法上,采用经济学实验的方法,设计了纳入过度自信的委托代理实验,并编制实验程序进行系统研究,分别检验拓展后的理论模型的最优解。在实验室的环境中,使控制CEO不同程度的过度自信成为可能,为理论模型提供了有效的实证支持,也为股东提供了更切合实际的参考和更深刻的理论依据。更多还原

【Abstract】 Taking the classic principle-agent theory and the relevant psychological theories as research basis, the overconfident psychological preference of CEO as research start point, the decision behaviors of shareholders and CEOs as research focus, the effort output risk, effort observability, and risk attitude as research main line, this dissertation constructs theoretical model with considering CEO’s overconfidence, and thoroughly probes into the problem of the designing CEO’s compensation contract for the shareholder, and then applies experimental methods to design economic experiments, in order to test the principal-agent model induced into CEO’s overconfidence psychology.Overconfidence is "the most steady finding in psychology", and one of the emotional factors and mental factors influencing economy man to decide. Although great of propositions have studied on overconfidence, how to test them using field data is still faced with obstacle, because overconfidence level can’t be evaluated accurately in real world. Based Berg et al (1986), by using experimental methods to control the relationship between trade units and probability, experimenter can control each preference function to induce participant to behave as their real preference. Based on the precondition that CEO is overconfident, this dissertation is established in extending the assumption of "homo economics" in classical principle-agent theory and constructs new theoretical model. Referring adequately researches of experimental economics in western countries and applying the computer experimental system, it controls laboratory conditions according to theoretical assumption and recruits participants to decide in fictitious principle-agent settings, and then tests the theory found on overconfidence by analyzing the experimental data.There are many factors influencing the principal-agent relationship. Choosing three vita factors of effort output risk, effort observability, and risk attitude from them as our research subjects, this dissertation focuses on how both principles and agents decide when they change. We extend our researches along with the consecution as follows: firstly, construct the principle-agent model between the shareholder and overconfident CEO; secondly, design and run experiments according to the theoretical propositions; finally, draw conclusions after analyzing the experimental data.This dissertation extends researches based on the above consecutions and draws the basic conclusions as follows: 1. Not only that subject’s actions in the repeated experiments convergence into the optimal action but also that the predicted optimal actions are significantly different from other non-optimal ones. The theoretical model is tested effectively and has some validity and significance in a certain extent. 2. Overconfidence has no significant relativity with risk attitude, man is more preferable risk and more overconfident. 3. The shareholders design different compensation contracts for different overconfident CEO when the conditions as risk level, moral hazard and CEO’s risk attitude vary.Based on the basic conclusion, there are following two contributions in this dissertation:1. In theory, the basic assumption of the classical principal-agent theory that CEO is rational is extended, and CEO is assumed to be overconfident here in chapter 3, 4, 5. Three factors of risk level, moral hazard and CEO’s risk attitude are considered into optimal compensation contract design for overconfident CEO under different conditions. It not only expands classical principle-agent model but also describes the relationship between shareholders and CEOs more clearly and actually.2. In method, experimental methods to address on principle-agent frame systematically through designing a series of economics experiments and making all experimental programs are used. The optimal solutions of extended theoretical models in chapter 3, 4, 5 are tested respectively. That makes it possible to control CEO’s different overconfidence level and provides effective empirical supports to theoretical models and offers more practical suggestions and more forceful academic evidences.更多还原