高分辨能力鲁棒一致性金融风险测度及其应用研究

Quality of financial risk measure largely determines the effectiveness of financial risk management. Selection of appropriate risk metrics and efficient calculation methods are the base to measure risk correctly. The project aims to establish a robust coherent risk measure with high resolution, and apply it to portfolio selection and capital asset pricing model. Firstly, from high-resolution and distributionally robustness perspective, we will re-examine that the desirable conditions and rules for risk measure, and expand the horizons for financial risk measure research. Secondly, according to the feature of probability distribution of assets’ return, we will construct some uncertainty distributed sets which fit the probability distribution and characteristic of assets’ return. Based on the conditions of high resolution, we will combine the development of rules and laws of risk measure. From the perspective of higher order stochastic dominant consistency and robustness, we aim at proposing a new class of high resolution and robust coherent risk measure. Thirdly, by analyzing the special structure of this class of risk measure, we will design fast and efficient stochastic optimization algorithms for it in order to provide new tools for measuring financial risk. Finally, we will apply it to portfolio selection model and capital asset pricing model, and make meaningful exploration for applications such as risk management and investment decision-making. Research on this subject, not only deepens our understanding about financial risk, but also promotes China's financial risk management in some degree.

金融风险测度的质量，很大程度上决定了金融风险管理的成效，而选择合适的风险度量指标和科学高效的计算方法是正确度量风险的基础。本项目拟建立高分辨能力鲁棒一致性风险测度，并将其应用于投资组合选择和资本资产定价问题。首先，从分辨能力和分布鲁棒性的视角，重新审视合意的风险测度应该具有的条件，拓展金融风险测度研究的视野。其次，根据资产收益的概率分布和特征，构建符合资产概率分布和收益特征的分布集。依据分辨能力条件，结合当前风险测度发展的准则和规律，建立一类新的具有高分辨能力的鲁棒一致性金融风险测度方法。再次，通过分析该类风险测度方法的特殊结构，设计快速高效随机优化算法，为金融风险测度问题提供新的计算工具。最后，将其应用于投资组合选择模型和资本资产定价模型，为其在风险管理和投资决策等应用做出有意义的探索。本项目的研究将深化对金融风险的认识，对于我国金融相关领域的风险管理水平也具有一定的推进作用。

随着经济全球化和金融一体化的发展，金融风险日趋复杂化和多样化。金融风险测度或定量方法作为金融风险管理的核心和基础，无论对投资人、金融机构本身、还是外部监管当局都至关重要。本项目建立了高分辨能力鲁棒一致性风险测度，进一步研究了资产配置和金融市场预测与定价等问题。该项目在计划研究期内完成了所有指标，主要获得了三部分重要成果。第一部分，根据资产收益的概率分布和特征，构建符合资产概率分布和收益特征的分布集，依据分辨能力条件，结合当前风险测度发展的准则和规律，建立一类新的具有高分辨能力的鲁棒一致性金融风险测度方法。第二部分是关于资产配置问题的研究。在收益-风险分析框架下，以高分辨能力的鲁棒一致性金融风险测为目标函数，考虑交易费用、均值收益限制约束、最小交易单位、正则化、模约束以及限制卖空等约束条件，建立了资产配置模型。进一步研究同标准情况下的投资组合选择效果的风险分辨能力检验，在不同持有期下考察投资组合的业绩，具体业绩指标采用投资组合夏普率、投资组合收益率和投资组合换手率等，验证资产配置模型的有效性。第三部分是金融市场预测与资产定价问题研究。从预测因子构建与选择，新的预测技术的应用和预测模型的改进等方面，显著提高了金融市场收益、波动率等的预测精度。遵循经典的Markowitz-Sharpe理论体系，金融市场驱动信息的有效性和高分辨率视角，建立资本资产定价模型，并将探讨两基金分离、有效边界、证券市场线等一系列问题。本项目的研究为全面认识金融风险提供了新的视角，对于我国金融相关领域的风险管理水平也具有一定的推进作用。