Non-Core Liabilities, Tail Dependence and Systemic Risks of the Chinese Banking Industry, a paper co-authored by our school’s Associate Professor Fang Yi, Associate Professor Jing Zhongbo from CUFE’s School of Management Science and Engineering, Wu Ji from New York University (an undergraduate of our school’s Outstanding Academic Talent Cultivation Program for 2019), and Associate Professor Li Zheng from the School of Finance of Tianjin University of Finance and Economics, was published in the 2020 4th Issue of the Journal of World Economy.
Starting from the business characteristics of banking institutions, the paper touches upon the accumulation process of the banking industry’s systemic risks and pinpoints the key sources of such risks by combining non-core liabilities with the tail dependence model which depicts correlation (like △CoVaR). And on that basis, it validates the effectiveness of the indicator through descriptive analysis and rigorous quantitative analysis. Specifically, the major contributions of this paper include:
First, establishing indicators (core indicators) that can effectively measure systemic risks of the Chinese banking industry by sorting through related theories and combining non-core liabilities with the tail dependence model to provide theoretical support to systemic risk prevention.
Through theoretical analysis, this paper holds that the use of data about the banking sector’s non-core liabilities can more accurately depict the accumulation of systemic risks on the time dimension than the traditional method of using stock-related data. This is because: ①From both the micro and macro perspective, non-core liabilities differ hugely from core liabilities, which subjects banking institutions to greater vulnerability when they assume a relatively high level of non-core liabilities, sowing the seeds of future risks. Nevertheless, data about the returns of bank stocks mainly measure the realization of risks, and may contain information beyond banks’ risks, which produces certain noises. ②Different from the fact that falling stock returns leads to a rising risk realization value, the rapid increase in non-core liabilities will result in a rise in risks accumulated by banking institutions. Risk realization and risk accumulation are two forms of risk, with the latter being more important (Fang Yi and Chen Min, 2019). Therefore, the paper studies systemic risks at the sources of risk.
Furthermore, after the ∆CoVaRindicator was put forward, many scholars have improved the estimation method for it through measurement techniques. In effect, ∆CoVaR can be improved along two lines: ①Around underlying data and the concepts of risk; ②Around correlation modeling techniques. Existing studies still use data of the stock markets and base their modelling on risk realization, so they work to improve the indicator along the ② line. But this paper seeks to improve ∆CoVaR along the ① line. In addition, it shifts from the traditional descending quantile dependence to ascending quantile dependence in establishing core indicators, which is more aligned with the idea of risk accumulation.
On the validation of empirical studies: ① Through the core indicators obtained from comparing the basic tail dependence model (△CoVaR) with the tail dependence model which gives more consideration to network correlation (Lasso-△CoVaR), it can be found that the performance of core indicators on the time dimension mainly depends on the selection of underlying data, other than correlation techniques or methods. ② The systemic risk indicators (core indicators) set up in the paper can effectively capture the volatility of the banking sector’s systemic risks amid typical shocks; for example, the four shocks during the sampling period include the 2008 global financial crisis (the first shock), the “money shortage” of China’s banking industry in 2013 (the second shock), unusual fluctuations of China’s stock market in 2015 (the third shock) and the China-US trade frictions in 2018 (the fourth shock).
Second, adopting multiple methods to validate the effectiveness of indicators. This paper comes up with three aspects of effectiveness for systemic indicators, i.e. effectiveness of capturing shocks, eliminating noises, and capturing scale characteristics. The first aspect of effectiveness means that the core indicators mentioned herein can effectively capture the shock of systemic risk events during the sampling period; the second means that the core indicators can eliminate information unrelated to banking businesses such as data about the stock market. The third means that the core indicators can accurately capture scale characteristics without additional scale factors and hence avoid the problem of “small institutions, big contributions”. This shows that the first two aspects target the systemic risks of the entire banking industry and validates the effectiveness of systemic risks on the time dimension. Whereas the third aspect targets banking institutions and validates the effectiveness of risk indicators on the spatial dimension.
During specific execution, the paper discusses the effectiveness of capturing shocks through analysis of the core indicator trend and the shock analysis method, the effectiveness of eliminating noises through placebo testing, and the effectiveness of capturing scale characteristics through the correlation between the ranking of systemically important banks and their scales obtained from relevant indicators.
Third, looking into the driving factors for systemic risks of the Chinese banking industry and providing theoretical support for the mitigation of such risks. The paper looks into the driving factors for the volatility of the banking sector’s systemic risks during the four shock events from the two perspectives of growth rate of non-core liabilities, and correlation of banking institutions. Then on such basis, it analyses the differences in the formation mechanism of the banking sector’s systemic risks during the four shocks.
Last, by increasing data frequency from quarter to month, and establishing core indicators with a scrolling window, the paper holds that the core indicators are affected little by the “forward-looking bias” when monitoring systemic risks.
