监测系统性金融风险——中国金融市场压力指数构建和状态识别

对系统性金融风险进行识别和评估,日益成为各国中央银行的核心关切。囿于数据频率,基于金融机构经营稳健性评估的金融系统性风险监测存在一定的滞后性,不利于中央银行及时进行风险应对,利用金融市场交易数据进行风险监测可极大程度克服滞后性问题。本文根据中国金融市场特点,选取债券市场、股票市场、货币市场和外汇市场17个有代表性的指标,运用经验累积分布函数法分别构造了各子市场的压力指数,以各子市场之间时变的相关关系刻画系统性金融风险的跨市场传染特征,合成金融市场压力指数,并通过建立马尔可夫区制转换模型,对金融市场压力状态进行识别。金融市场压力指数能有效反映样本域内的压力事件,并兼具稳健性、能逐日监测等优点,为监测评估系统性金融风险、选择政策实施窗口和评估政策实施效果等提供了有力工具。

Monitoring Systemic Financial Risks:Construction and State Identification of China's Financial Market Stress Index

For the central bank to maintain financial stability and carry out macro-prudential management, it is essential to have timely and efficient monitoring of financial market conditions. The stability of financial institutions depends on the conditions of the financial market, and the effects of monetary policy and macro-prudential policy are transmitted through the financial market; the policies themselves are also responses to financial market conditions. In addition, financial market data contain highly forward-looking information. Major changes in the financial and economic system, such as policy adjustments and stress events, will be reflected in the financial market data. The central bank also needs to closely monitor financial market conditions to select the policy implementation window in advance, make adjustments during policy implementation, and evaluate the policy effect. A good method of monitoring the overall risk level of the financial market is to construct a financial market stress index with selected indicators of the financial market. Overseas researchers and institutions, and more recently domestic researchers, have extensively explored the construction of financial market stress indexes. Most financial market stress indexes constructed by domestic researchers can identify financial market stress events, but the index construction and stress state identification still show deficiencies. The frequencies of financial market stress indexes in the literature are relatively low, as they are limited by data availability and construction methods. Some studies use indicators such as the non-performing loan ratio of the banking sector, but the data have some lag and can be manipulated. We believe that constructing a financial market stress index with pure financial market data can address the deficiencies of the literature. Furthermore, as interest rate liberalization continues, the representativeness and effectiveness of a financial market stress index that measures systemic financial risk using financial market data will be further improved. In this paper, the construction of the financial market stress index involves two steps. The first is the construction of each sub-market stress index, and the second involves compiling the full financial market stress index based on these sub-market stress indexes. This paper selects 17 indicators, calculated with transaction data from China's bond market, stock market, money market, and foreign exchange market, to construct sub-market stress indexes using the empirical cumulative distribution function method. It then constructs the financial market stress index with the sub-market stress indexes, using the time-varying correlation between them to depict the cross-market contagion characteristics of systemic financial risk. The purpose of constructing the financial market stress index is to monitor and evaluate the stress level of the financial market, especially high stress states. Some studies define a high stress state as occurring when the current value of the financial market stress index exceeds the mean of its historical values by a specified number of standard deviations. Other studies determine stress states by comparing the current value of the financial stress index with its values during the financial crisis. None of these methods make sufficient use of the information contained in the financial market stress index. The Markov regime switching model proposed by Hamilton is a more proper method for identifying financial market stress states. This paper assumes that there are two stress states in the financial market—high and medium-to-low, which is preliminarily supported by the analysis of the historical distribution of the financial market stress index. It then establishes the Markov regime switching model to identify stress states. Through back testing, our financial market stress index is found to accurately reflect historical stress events; for example, the large number of securities firms on the verge of bankruptcy in 2003, the global financial crisis of 2007-2008, the European sovereign debt crisis, interbank liquidity strains in June 2013, abnormal stock market fluctuations in 2015, and the COVID-19 outbreak. Our financial market stress index, which has the advantages of robustness and high frequency, is a powerful tool to monitor and evaluate systemic financial risk, select a policy implementation window, and evaluate the policy effect.