【作者】 黄娟娟；

【导师】 沈艺峰；

【作者基本信息】 厦门大学 ， 企业管理， 2009， 博士

【副题名】基于我国上市公司股利“群聚”现象的研究

【摘要】 股利政策是财务学研究的经典课题之一,也是现代财务理论历经半个多世纪的发展而未能最终解答的研究难题。传统财务学关于股利政策的研究从“有关论”到“无关论”,而后开始逐步放宽MM定理的假设,产生了“追随者效应”、“信号传递”和“委托代理”三大理论,但不同的学说各执己见,在揭示了股利政策的某方面特征的同时,也留下许多的疑点,迄今为止没有一个理论能够完整地回答所有的“股利之谜”。一方面,财务学家重新审视MM定理时发现,传统股利政策陆续放宽了无税收、信息对称、完美代理人等多项假设,但从未触及过理性人假设,忽略了人的行为偏差对股利政策可能产生的影响。Shiller(1984)指出,除非未来的模型考虑了这些影响,否则人们对股利的非理性偏好很难解释,Miller(1986)也认为,在大多数微观决策的蓝图上,行为的因素是非常重要的组成部分,它不应该被忽略,把行为因素引入股利政策的解释有助于解释长期以来的异象。另一方面,股利支付作为资源分配行为,必然存在需求方和供给方。类似追随者效应的传统股利政策都只考虑了股利的需求方,或需求方对支付行为的主观解读,而作为股利提供方的管理者行为背后的主观能动性及其形成原因反而被忽视了。因此,有必要从股利供给方的角度对股利支付进行研究(Baker和Wurger,2004a)近年来,随着实验心理学和行为学的发展,人的行为因素开始被引入财务学领域的研究,行为公司财务的研究逐渐兴起,而从管理者行为的角度研究股利政策也成为当前股利政策研究的前沿课题。2003年,Gurtler和Hartmann的工作论文《行为股利政策》首次提出了“行为股利政策”一词,使之正式登上了股利政策研究的舞台。但遗憾的是,Gurtler和Hartmann(2003)虽然在标题重要位置提出了“行为股利政策”,但文中并未对它的内涵和外延进行研究和拓展,只是指出行为股利政策是行为公司财务的一个组成部分。尽管此后一些有影响的成果(例如Baker和Wurger,2004a、2004b;Li和Lie,2005)从行为角度探讨股利政策,但是这些零散的文献也仅就自身角度进行分析,迄今为止尚未有文献将它们概括为“行为股利政策”,或者专门针对“行为股利政策”这一课题展开较为全面和系统的研究,而且相关实证分析也没有充分展开,因而经验证据也很少。因此,本文就以“行为股利政策”为题进行相关研究。本文研究行为股利政策的另一个原因在于,本文发现了一个之前股利政策的研究从未提及的新异象——我国上市公司的股利“群聚”现象。我们发现,80%以上的上市公司每股现金股利在0.2元以下,以变异系数表示的股利分散程度只相当于同时期美国上市公司每股现金股利分散程度的三分之一。为何会产生股利“群聚”现象?本文认为,传统股利政策理论无法提出合理的解释,而引入行为因素则有助于分析这些异象产生的原因(Shefrin和Statman,1984)。因此,本文通过探讨管理者的行为对公司股利政策的影响,以“行为股利政策”为工具来解释中国上市公司股利“群聚”现象。本文的研究内容主要包括文献综述、制度背景分析和经验研究,全文共分为八章,各章的具体内容如下:第一章为导论,介绍本文的研究问题、研究内容、论文框架、研究方法以及主要学术贡献与创新。第二章对传统股利政策理论的研究文献进行回顾。本章首先回顾了完美市场上股利无关论,然后介绍了现代股利政策理论如何逐渐放宽MM定理的某些假设,形成了股利的税收、信号和委托代理三大理论,最后总结和评述传统股利政策理论所面临的挑战。第三章描述我国上市公司存在的股利“群聚”现象。本章首先提出股利“群聚”现象,然后指出“群聚”现象存在于各个年度,而且具有行业分布不同的特征。最后本章通过分析指出,西方传统股利政策理论无法解释我国上市公司的股利“群聚”现象。第四章为行为股利政策的文献综述。本章首先介绍了行为股利政策理论的产生过程,指出目前行为股利政策可能的两个发展方向为股利迎合理论和股利羊群行为理论。然后,本章分别从股利迎合理论和管理者羊群行为两个角度进行文献评述,为后续的实证研究提供理论支持。第五章是管理者羊群行为与股利“群聚”现象的实证分析。本章的实证检验分成两个部分,第一部分检验了股利羊群行为是否存在,第二部分检验了股利羊群行为产生的原因是否与声誉有关。第六章是股利迎合理论与股利“群聚”现象的实证分析。本章的实证检验同样分成两个部分,第一部分检验了我国上市公司管理者支付股利时是否迎合流通股东的偏好,第二部分检验了管理者支付股利是否迎合了大股东的现金偏好。第七章是行为股利政策的联合检验。本章采用Tobit模型,把管理者股利羊群行为和股利迎合理论放在同一个模型内联合检验,同时考察“是否支付”与“支付水平高低”问题以克服可能存在的样本选择偏差。本章也是行为股利政策及其对“群聚”现象解释能力的稳健性检验。第八章为全文总结,具体包括研究的主要结论与启示,研究的局限性及未来的研究方向。本文的主要研究结论如下:第一,管理者的股利羊群行为可以解释我国上市公司股利“群聚”现象,一方面,管理者在决定股利支付水平时受到行业普遍水平和行业领先者的影响而产生了羊群行为。另一方面,公司的声誉是产生和影响羊群行为程度的动因,声誉越高的公司越不可能产生羊群行为。第二,管理者股利迎合行为同样可以解释我国上市公司股利“群聚”现象。管理者支付股利时没有考虑中小股东的股利需求,而是仅仅迎合了大股东的现金偏好。上市公司股权越集中,越倾向于支付股利,支付的金额也越高。最后,大股东偏好也与我国上市公司股利的“群聚”程度也存在正相关关系。第三,采用Tobit模型,对行为股利政策理论进行联合检验发现,管理者羊群行为和管理者迎合大股东行为对上市公司的股利政策有稳定的影响,公司声誉和股权集中度对均股利“群聚”存在稳定的相关关系。总之,本文的研究结果证实,管理者的羊群行为和迎合行为在很大程度上解释了我国上市公司的股利支付行为,对声誉和大股东偏好的考虑是管理者支付行为的动因,因此,行为因素是影响股利支付的重要因素,行为股利政策能够解释我国上市公司的股利“群聚”现象。本文的主要贡献与创新之处表现在如下三个方面:在选题上,首次提出我国上市公司股利支付存在的“群聚”现象,丰富了对“股利之谜”的研究;在文献评述上,较为系统和全面地梳理了管理者羊群行为和管理者迎合理论的相关文献,首次将管理者羊群行为理论应用于股利政策领域,并首次将管理者羊群行为和迎合理论归纳和总结为行为股利政策理论;在实证研究上,采用Tobit模型检验股利“群聚”现象,克服仅采用线性回归模型研究股利问题可能存在的样本选择偏差的问题。本文的局限性在于,仅仅从经验研究的角度探讨行为股利政策对股利“群聚”现象的影响,未能进一步构建一个正式模型来描述管理者行为因素,特别是管理者羊群行为因素对股利政策的影响。更多还原

【Abstract】 Being one of the most classical fields in science of finance, dividend policy is still a much-debated problem yet to be solved over the last five decades. Traditional finance research on dividend started with the debate between the dividend irrelevance and relevance propositions. Afterwards researchers turned to gradually relax the assumptions of MM theorem and three important theories on dividend, the clientele effect, the signaling models and agency theory, were thereof put forward. Despite the distribution on unveiling dividend policy in various aspects, those theories persist in their own opinions and many doubtful points left behind. And so far none of them could explain the whole puzzle of dividend. On the one hand, when reviewing the studies on MM theorem, researchers found that several assumptions were relaxed such as no taxes, information symmetry and perfect agency assumption, but the "rational man" assumption remains untouched upon and the possible impacts of human behavior bias on dividend were ignored by those studies. Shiller (1984) pointed out that it is very difficult to explain the irrational preference to dividend by investors unless aforesaid impacts were taken into consideration in future models. Miller (1986) agreed that being an important part of most micro decision making, behavioral elements could not be ignored and the introduction of behavioral elements in dividend policy might help resolve longstanding anomalies. On the other hand, as a method of resource allocation, the payment of dividend has two sides: supply and demand. Traditional research on dividend policy, such as the clientele effect, only considers the demand side or the subjective analysis on the payment behavior by the demand side and ignores the manager behavior and the incentives behind it as the supply side. Therefore, it is necessary to study dividend payment from the standpoint of the supply side (Baker and Wurger, 2004a) .In recent years, with the development of experimental psychology and behavioral science, the behavioral elements have been introduced into finance study. Behavior corporate finance has been a rising field and the study in the sight of manager behavior is becoming one of the frontier topics in the field of dividend policy. In 2003, Gurtler and Hartmann put forward the term "behavior dividend policy" in their working paper, the behavior dividend policy, which is the first time the term stepped onto the stage of dividend research. But it’s a pity they only point out that behavior dividend policy is a part of behavior corporate finance and actually they don’t elaborate and define the concept’s connotation and extension. Some influential works afterwards, such as Bake and Wurger (2004a, 2004b), Li and Lie (2005), studied dividend policy from the standpoint of behavior. But they are unsystematic and isolated from each other. Until now there is no research which classifies those works under the field of behavior dividend policy and no systematic research focusing on the field of behavior dividend policy. The empirical research in this field was underdeveloped therefor the empirical evidences are rare. The paper aims to do some research focusing on the behavior dividend policy.Another reason for choosing this topic is that a new anomaly which is never mentioned before in the works on dividend policy was found in this paper - the dividend clustering phenomena in the listed companies of China. We found that 80% of the listed companies paid cash dividend under 0.2 RMB Yuan per share. The dispersion of dividend in China showed in coefficient of variation only equals to 1/3 of that in America. What is the origin of the dividend clustering phenomena? Since the traditional dividend policy theories can not explain this phenomena, it will be helpful for us to introduce behavioral elements to analyse the cause of the phenomena (Shefrin and Statman, 1984). So, using "behavior dividend policy" as a tool, the paper is intended to explain the dividend clustering phenomena in listed companies of China by discussing the impacts from manager behavior to corporate dividend policy.The paper begins with a literature review and then followed by background analysis and empirical research. It is divided into eight chapters.Chapter one is an introduction of the research issues, contents, framework, methods and the contributions of the paper.Chapter two is literature review of researches on traditional dividend policy. This chapter reviews the dividend irrelevance propositions in perfect market and then introduces how the assumptions of MM theorem were relaxed by theories of modern dividend policy so as to form taxes clienteles, signal models and agency hypothesis. Finally, a summary and discussion of the challenges that the traditional dividend policy theories face will be made.Chapter three shows the phenomena of dividend clustering of listed companies in China. This chapter firstly points out that the dividend clustering phenomena happens in every year and its distribution has various characters in different industries. After analysis the chapter concludes that traditional research on dividend policy in west countries can not explain the dividend clustering phenomena in China.Chapter four is a literature review of behavior dividend policy. The chapter shows how the behavior dividend policy was formed and then points out two possible branches developing from it - the catering theory and the herding theory of dividend. After that, the chapter reviews the literature of two branches separately, which is providing the theoretical foundation of empirical research thereof.Chapter five is empirical research on relationship between herding behavior of managers and dividend clustering. There are two parts in the chapter: part one examines whether the herding behavior of managers exists in dividend payment, and part two discusses if the herding behavior of dividend payment relates to the reputation of the company.Chapter six is empirical research on relationship between catering behavior of managers and dividend clustering. There are also two parts in the chapter: part one studies if the managers of listed companies in China cater to the preference of tradable shareholders in dividend payment and part two shows if the managers cater to the cash preference of leading (untradeable) shareholders.Chapter seven is a joint examination of behavior dividend policy. Tobit model was used in this chapter to examine both herding and catering theories in one single model and to study "whether to pay" and "how much to pay" at the same time. This chapter is also a robustness testing of explanation of dividend clustering phenomena by behavior dividend policy.Chapter eight is the conclusion of this paper. It sonsists of the research finding, the lessons we learn, the limitations of the research and the directions for future research.This paper has reached the following conclusions:Firstly, herding behavior of managers can explain the dividend clustering phenomena of listed companies in China. On one hand, the payment decisions of manager will be influenced by the last average dividend level of leading companies and the last average dividend level of all the companies in the industry. On the other hand, company reputation is a factor which can influence herding behavior. The higher the reputation is, the lower probability in which herding behavior happens.Secondly, catering behavior of managers can also explain the dividend clustering phenomena of listed companies in China. Managers in China do not consider the preference of small and middle shareholders. They only cater to the cash demand of leading shareholders. The more shares concentrated, the more likely that managers turn to pay and at the same time the amount of dividend paid is higher. What’s more, leading shareholders’ preference is positive related to the level of concentration of the dividend for the listed companies in China. That’s one of the reasons why the dividend clustering phenomena will happen.Thirdly, using a Tobit model to examine the behavior dividend policy theories, the paper finds that both herding and catering behavior of managers can explain why the dividend will cluster. The company reputation and shares concentration will influence dividend clustering positively.In a word, this paper offers evidences that herding and catering behavior of managers can explain the dividend behavior of listed companies in China. The consideration to company reputation and leading shareholders’ preference are incentives of the dividend payment. As a result, the behavioral elements are important factors to influence dividend payment, and the theory of behavior dividend policy can explain the dividend clustering phenomena of listed companies in China.The major contributions and innovations of this paper are in the following aspects: in subject selection, the paper is a precursor to study the dividend clustering phenomena of listed companies in China, which enriches the research on the dividend puzzle. In literature review, this paper is the first work that applies herding behavior of managers in the area of dividend and integrates the literatures of the herding and catering behavior of managers under the theories on behavior dividend policy. In the empirical study, this paper applies the Tobit model to examine dividend clustering phenomena, which can reduce the risk of sample selection bias that might happens in linear regression models.This paper discusses the impact on dividend clustering phenomena of behavior dividend policy from the single perspective of empirical research. But, for the aspect of building a formal model to elaborate manager behavior especially the impacts on dividend policy from herding behavior of managers, much progress needs yet to be made.更多还原