跨市场风险传递的非对称多重分形DCCA方法构建及其应用研究

Constructing the asymmetric multifractal DCCA method can enrich time series research methods, on which empirical research on China's financial markets based may provide decision-making reference for China's macro-financial regulation and micro-portfolio. In our project, three types of asymmetric multifractal DCCA methods are proposed in connection with different kinds of trends, fluctuations and transfer directions for pairwise time series. By using multivariable ARFIMA and MFbs models, the methods of generating artificial numerical data are constructed, and by means of simulation the effectiveness and credibility of asymmetric multifractal DCCA method are tested. Furthermore, the relationship between the sample size and method credibilty and the effect of different detrended linear and nolinear filtering on asymmetric multifractal DCCA method are also discussed. Thenm, the asymmetric cross-market risks associated with China's stock market and their multifractal charactistics are empirically analyzed, to illustrate the financial risk transfer mechanism. Additionally, through the sliding window technique, the time-varying characteristics and the effect of the major events on this mechanism are analyzed in-depth. Finally, from the views of the data distribution structure and economiv principles, the reasons of asymmetric multifractal characteristics of the analysed cross-market risk transfer are discussed, on which the financial regulation policy suggestions are given based.

非对称多重分形DCCA方法的构建有助于丰富时间序列研究方法,其针对我国金融市场的实证研究可为我国宏观金融调控和微观投资组合提供决策参考。本课题将针对成对时间序列，构建不同趋势、不同波动幅度、不同传递方向的三类非对称多重分形DCCA方法，并借助多元ARFIMA和MFbs等模型，提出生成此三类特征的时间序列的人工数值模拟方法，通过仿真模拟来验证非对称多重分形DCCA方法的有效性和可信度、样本容量与可信度之间的关系，以及不同消除趋势的线性和非线性滤波对非对称多重分形DCCA方法的影响。在此基础上，对与我国股市相关的跨市场风险的非对称相关性及其多重分形特征进行实证分析，以说明国内外金融市场对我国股市的风险传递机制，并借助滑动窗技术，深入分析这种机制的时变特征和重大事件对此机制的影响情况。最后，从数据分布结构和经济原理两个角度给出我国跨市场风险传递非对称多重分形特征的原因，并给出金融调控的政策建议。

本课题严格执行项目计划书的要求，在国内外期刊共发表相关学术论文22篇，Springer出版社出版著作1部，准时高效的完成了研究计划。.本课题针对两个时间序列，从三个角度（不同趋势、不同波动幅度、不同传导方向），两类不同消除趋势方法（线性和非线性），通过非对称多重分形DCCA 方法的构建和仿真模拟来对界定方法的有效性和可靠性。在MF-DCCA、A-DFA等方法基础上，构建了针对上涨和下降的不同趋势的非对称多重分形DCCA方法（MF-ADCCA）、针对不同波动幅度的非对称多重分形DCCA方法（VC-MF-DCCA）、针对不同传导方向的非对称多重分形DCCA方法（DMF-ADCCA）。另外，利用线性和非线性的滤波消除趋势：经验分解模型（EMD）和极大重叠的连续小波变换（MODWT）分别构建了MFDFA-EMD方法和MFDCCA-MODWT方法，并对这些方法进行了仿真模拟，丰富了多重分形消除趋势类方法。在此基础上，对与我国股市相关的跨市场金融风险的非对称相关性和多重分形特征进行实证分析，发现中国股市与国外主要股市、国内汇市、国内股指期货市场间皆存在长记忆特征，且其交叉相关性呈现出非对称的多重分形特征。国内外股票市场的价格波动存在双向传导效应，且从国外市场向国内传导的影响程度更大，而股票市场的波动来源于市场发生的重大事件。实证结果对于构建跨市场资产组合具有决策参考意义。