全球股票市场间风险传染的测度、监管及预警

近年来,伴随金融一体化程度的加深,全球各股票市场间风险传染的动态复杂性加剧,其准确测度、高效监管及实时预警已成为优先事项。本研究选取全球21个代表性股票市场作为分析样本,首先基于广义向量自回归模型的滚动估计准确测度其间风险动态传染的高维网络序列,进一步借由矩阵值因子模型来稳健收缩上述序列,以探究其潜在动态核心结构,从而实现高效监管。最后,通过向量自回归模型的预测功能实现对全球股票市场间风险传染的实时预警。研究表明,全球股票市场间风险传染具有时变性,其监管与预警可通过少数与地理区域高度相关的风险区域间的动态传染关系及内部的市场构成来刻画。与此同时,我们发现中国内地等新兴市场的重要地位逐渐凸显。本文研究结论可为有效防范与化解金融风险提供有益参考。

Measurement,Supervision and Early Warning of Risk Contagionamong Global Stock Markets

Economic globalization and financial integration have increased, strengthening the network effect of global financial markets and the resonance of market risks. Researchers should no longer ignore risk contagion because it is an important component for understanding the financial markets. The stability of the stock market no longer depends on its individual volatility, as it is now vulnerable to spillover from other markets. Risk contagion in global stock markets is difficult to research because the close relationships between multiple market entities in the international stock market system complicate the data structure of risk contagion's (high-dimensional matrix-value time series). In addition, the risk management of the international stock market system will become more difficult as emerging stock markets gain international recognition. The financial regulators will face two problematic choices, “too big to fail” and “too interconnected to fail.” Therefore, researchers have begun to prioritize accurate measurement, efficient supervision, and real-time early warning signs of risk contagion in global stock markets. Researchers have also begun to study the potential core structure of risk contagion.
This study selects 21 stock markets from four geographical regions, Asia, Oceania, Europe, and the Americas, as its sample data. Firstly, we construct a high-dimensional matrix-value time series based on the generalized vector autoregressive model to investigate the dynamic network effect of risk contagion. The time series represents the size and direction of risk contagion among global stock markets. Secondly, considering the widespread existence of financial data outliers, we use the high-dimensional matrix-valued factor model's robust dimensionality reduction function to extract potential risk communities and identify the dynamic core structure of risk contagion between global stock markets. This provides efficient supervision. Thirdly, we use the vector autoregressive model's prediction function to identify the real-time early warning signs of risk contagion's core structure between global stock markets in the next six months.
The empirical results show a time-varying pattern of risk contagion among global stock markets. Although the patterns are time-dependent, three risk communities can always be identified as early warning signs. The contagion relationship between and within the three risk communities describes the dynamic core structure of risk contagion in global stock markets. The three risk communities have strong geographical attributes. This study's empirical conclusion will improve the concept of real-time hierarchical risk management, as the findings demonstrate that the risk management of the international stock market system must be divided into two steps: firstly, dynamic monitoring risk contagion among a small number of communities to identify the main path of risk contagion among global stock markets; secondly, use the regional characteristics of each risk community to implement real-time risk management. In addition, this study provides policy recommendations regarding the role of risk-contagion in emerging stock markets and the idea of “inter-regional and within-regional” risk governance.
The study makes the following contributions. Firstly, the high-dimensional matrix-valued factor model is introduced to the study of risk contagion in global stock markets. The study improves the model's robust estimation to effectively reveal risk contagion's dynamic core structure, expand the model's scope, and provide new opportunities for a follow-up study on financial risk contagion. Secondly, the paper utilizes the model proposed by Diebold and Yilmaz (2012) to analyze the time-varying volatility spillover effect of 21 developed and emerging stock markets from four geographical regions. The model and sample data provides a more comprehensive and clear understanding of geographical relationships and financial risk. Lastly, based on the identified dynamic core contagion relationship between a small number of risk communities and their market composition, the real-time hierarchical risk management concept proposed in this study can provide a useful reference and method of supervision for the real-time early warning signs of risk contagion in international stock markets.