基于广义线性函数的随机占优统计推断与证券市场投资者总体偏好

Stochastic dominance approach is applied to various areas of risk decision. The examinations of the approach can be divided into three types: pair-wise dominance relations, convex dominance relations and portfolio efficiency dominance relations. The tests of pair-wise dominance relations have been well developed. However, tests of convex dominance relations and portfolio efficiency relations need further improvements. The project intends to study on statistical inference of stochastic dominance based on latest method of generalized linear formulations （Post and Kopa, 2012): firstly, develop the asymptotic and bootstrap statistical inference for the portfolio efficiency dominance relations; secondly, subdivide the relations of convex dominance, and extend different statistical inference methods of pair-wise dominance relations and portfolio efficiency dominance relations to the subdivided relations on the basis of the identical and distinguish between the subdivided relations and pair-wise dominance relations and portfolio efficiency dominance relations. Moreover, the project focuses on market efficiency and equilibrium, and investigates mean-variance efficiency and many aggregate preferences assumptions of economics and finance related to financial market anomalies and puzzles in various markets to explore general laws and special rules. In addition, the project will synthesize the findings of examinations of pricing kernel, pricing error and aggregate preferences to provide some important explanations for economic and financial theories and models, and give valuable advice on practices of asset pricing, portfolio management and risk management.

随机占优方法被应用于风险决策各个领域，其检验可分为成对占优关系、凸占优关系和组合效率占优关系检验。成对关系检验较成熟，而凸占优和组合效率占优关系检验仍有许多不完善之处。本项目拟对随机占优的最新广义线性函数方法(Post和Kopa，2012)的统计推断加以研究：第一，提出组合效率占优关系检验的渐近和Bootstrap统计推断；第二，对凸占优关系进行进一步细分，考察细分的不同情况下检验统计量特征，利用它们与成对占优和组合效率占优关系检验统计量的联系与区别，将相应的统计推断方法推广到凸占优关系的统计推断。同时，本项目从市场效率和市场均衡出发，针对金融市场"异象"，对不同证券市场，进行均值方差效率，多种经济金融学总体偏好假定的检验，探讨市场的一般和特殊规律。另外，将定价核、定价误差与总体偏好检验结果相结合，对经济金融理论和模型给出一些重要解释，对资产定价，组合管理和风险管理的实务给出有价值的建议。

课题围绕广义线性函数的随机占优(Stochastic Dominance, SD)检验方法和投资者总体偏好进行研究，完成了计划书研究内容,并增加了以下研究：第一，高阶SD准则充要条件；第二，“适合风险厌恶”的线性SD准则；第三,市场投资者偏好的应用。课题发表论文6篇，其中SSCI和SCI论文3篇。SD方法被应用于风险决策各领域，总体偏好是建立经济金融学和管理学理论的基石。本课题通过建立高阶SD准则及其统计推断方法，并检验市场投资者总体偏好，是对近年Journal of Finance，Review of Financial Studies, Management Science和European Journal of Operational Research等国际顶级期刊中相关前沿研究的延伸和拓展。主要贡献：第一，修正广义线性函数的高阶矩SD准则，提出了效率占优和凸占优关系高阶矩SD充要条件；第二，给出了具有“适合风险厌恶”的DARA的凸占优和组合效率占优的线性检验方法，这一工作开创性的建立了非线性风险偏好约束的线性SD检验方法，提供了更深刻刻画投资者风险决策的有力工具；第三，在Hall和Horowitz(1996)的基础上，结合Barrett和Donald(2003)和Linton等(2005)，提出了具有一致性的Bootstrap检验；第四，通过对不同市场“异象”的Benchmark 组合对投资者的总体偏好研究发现，采用二阶SD进行投资分析会缺乏判断力，高阶SD规则对经典MV分析具有重要的完善作用，而且被动投资的市场指数并不有效，市场存在基于高阶风险偏好的套利机会。另外，课题在基金经理特质、基金绩效、机构投资者等方面也取得了一些成果；专栏作者Simon专门为课题基金经理研究的论文撰写评论，发表在华尔街日报的门户网站MarketWatch，并被全球众多著名财经网站转载。