数据驱动下资产泡沫的识别与明斯基时刻的预警研究

The ecological environment of China's financial markets has become increasingly complex. Market volatility and risk contagion caused by uncertainties often put the entire financial system in the edge of systematic failure. Effectively identifying various types of asset bubbles in a complex and uncertain environment, and using early warning indicators to dynamically monitor the accumulation of major risks caused by the bubbles, are of great strategic significance for improving China's macro-prudential management system and steadily advancing the financial services innovation. Along with our country's strategic deployment and from the theoretical perspective of a complex system, this project elucidates the evolutionary patterns, micro foundations, and mechanisms of different types of asset bubbles (such as for stocks, real estates and digital currencies) by integrating statistical analysis results, empirical results of behavioral finance and simulation techniques of computational finance. And then the theoretical basis for identifying asset bubbles will be provided. Moreover, by taking a data-driven approach to build a distributed modeling framework for dynamic back-testing and prediction, empirical tests on the major historical crisis events can be further conducted. Data mining, pattern recognition and multi-scale analysis methods are applied to fuse multi-feature at different levels, to construct early warning indicators of Minsky moments that are more robust and suitable for China. Furthermore, the decision support for the reasonable selection of macro-prudential policy tools is provided.

中国金融市场的生态环境日趋复杂，由不确定性因素引发的市场大幅波动及风险传染常使整个金融体系面临着发生系统性失灵的可能性。如何在复杂不确定环境下有效识别各类资产泡沫，并利用预警指标动态监测由资产泡沫化引起的重大风险累积，对完善我国宏观审慎管理制度并稳步推进金融服务创新有重大的战略意义。为响应国家的战略部署，本项目基于复杂系统的理论视角，通过集成统计分析的结果、行为金融学的实证结果、计算金融学的仿真技术，阐明股票、房地产和数字货币等资产泡沫的演化模式、微观基础和作用机制，进而给出识别资产泡沫的理论依据；以数据为驱动，搭建由异质微观主体交互作用而涌现出的以分布为对象的、可实现动态回测和预测的框架模型，并对历史上重大危机事件进行实证分析，再利用数据挖掘、模式识别和多标度分析等方法融合不同层级的特征信息，以构建适合我国且更为稳健的明斯基时刻的预警指标；进而为合理选择宏观审慎政策工具提供决策支持。

中国金融市场的生态环境日趋复杂，由不确定性因素引发的市场大幅波动及风险传染常使整个金融体系面临着发生系统性失灵的可能性。如何在复杂不确定环境下有效识别各类资产泡沫，并利用预警指标动态监测由资产泡沫化引起的重大风险累积，对完善我国宏观审慎管理制度并稳步推进金融服务创新有重大的战略意义。本项目基于复杂系统的理论视角，通过集成统计分析的结果，阐明资产价格时间序列的演化模式，进而给出识别资产泡沫的理论依据；面向多源异构数据，搭建可进行信息融合、可实现动态回测和预测的学习框架，通过比较成熟的和新兴的两类金融市场，进一步检验信息传导至资产价格的有效性及资产价格的可预测性，丰富对“新兴市场情报假说”的理解；利用数据挖掘、模式识别和多标度分析等方法融合不同层级的特征信息，以构建适合我国且更为稳健的明斯基时刻的预警指标，进而为合理选择宏观审慎政策工具提供决策支持。