基于临界型泡沫理论的系统性风险测度方法研究

The subprime crisis and its repercussions have brought macro-prudential regulation back to the international top focus for coping with future possible financial crisis. The utmost importance issue of taking macro-prudential regulation is measuring the systemic risk. From perspective of complex financial system, the primary characteristic of systemic risk can be regarded as the endogenous emergence of global instability during market operation, which stems from the spontaneous divergence for length of interaction connections within trading network. The theory of financial bubbles with critical times is developed to predict impending market crash by diagonsing degree of trading coupling for whole market and expectation synchronization for traders. This theory naturally provides a suit of practical and effective analytics to capture the instability of financial system. Therefore, it is feasible to develop some measures of systemic risk that have crisis-warning effects via further extending such bubble theory. In beginning of this project, a quite important problem, which is the estimation of time-varying critical times, is solved under a special state-space model by particle filter . Then, the estimated critical times are use to construct four key indicators regarding systemic risk issues including tail risks, market susceptibility, distribution of systemic risk contributions and exposure to systemic illiquidity. This part of research is carried out by a variety of statistical techniques, such as Monte Carlo simulation for stochastic differential equation with jumping term, analysis of Granger-causal network, multivariate econometrics, etc. In the end, a comprehensive approach to measure systemic risk and applied to macro-prudential regulation is integrated and refined by conducting respectively the time series and the cross-sectional empirical tests for Chinese financial system, which might provide evidence and policy proposal for Chinese regulator coping with financial crisis. This project is of positive significance for supplementing method system of measuring systemic risks and making crisis warning.

次贷危机之后，宏观审慎监管重新成为应对金融危机的国际焦点问题，首当其冲的就是系统性风险的测度。本项目认为，系统性风险的首要特征是金融运行内生涌现的全局不稳定性，而临界型泡沫理论通过甄别交易耦合度和预期同步性的异常演化来预警泡沫破裂，实际上提供了一套捕捉系统全局不稳定性的实用方法。因此，本项目考虑对该理论深入拓展来开发具备预警功效的系统性风险测度方法。本项目首先解决临界型泡沫理论的一个重要的科学应用问题，在状态空间系统下实证估测出时变临界时点，并对模型族进行筛选。然后，运用随机微分方程的蒙特卡洛模拟、格兰杰因果网络、多元计量等方法，围绕估计出的时变临界时点分别构建尾部风险、市场敏感性、险位分布和系统流动性四类关键测度指标。最后，在对中国市场时间维和空间维上的实证基础上，统合出适用于我国宏观审慎监管的系统性风险测度方法。本研究对于系统性风险测度方法体系的补充和危机的预警具有积极意义。

系统性风险的首要特征是金融市场内生涌现出的全局不稳定性，临界型泡沫理论提供了捕捉系统全局不稳定性的实用思路。本项目围绕对临界型泡沫模型构建、求解估计和衍生拓新等方面的研究，来探索开发具有预警功能的系统性风险测度指标。模型构建上，研究了临界型泡沫的随机折现因子构建，连贯解释风险积聚与释放；研究了Markov区制转换的随机临界时点泡沫模型，内生建模短期投机涨落；研究了动量交易视野时变的临界型泡沫模型，在暂态自验性框架下刻画预期同步性反转；求解估计上，开发了基于隐Markov链的期望最大化估计和Lamperti变化下的拟似然估计两种方法；衍生拓展方面，研究了中国的随机尖点突变泡沫理论和基于定价核还原原理的临界型泡沫诊断方法。基于以上研究，本项目提出一套监测系统性风险内源时空关联的研究工具：投机影响网络(SIN)。通过SIN，研究了市场尾部风险、敏感性、险位分布和流动性四类指标的计算。同时，本项目针对中国股票市场、基金市场和房地产市场进行了实证，探索出一些预警效果的风险测度值。本项目完成标注论文6篇，其中4篇已发表，1篇已接收，有两篇在SSCI期刊中发表。