Journal of Risk Model Validation

Steve Satchell
Trinity College, University of Cambridge

This issue focuses somewhat unintentionally on Chinese banks, while also looking at bank risk. However, these papers have primarily been selected for their contributions to risk validation, and it seems that much of the cutting-edge research in this area is emanating from Asia. We are delighted to publish these excellent papers in The Journal of Risk Model Validation.

Our first paper, “Measuring the systemic importance of Chinese banks: a comparison of different risk measurement models”, is by Chunlin Cai, who uses a number of well-known risk measurement models to measure systemic risk and then compares their performance in measuring the systemic importance of banks. His results show that the different risk measurement models yield significant differences in systemic risk. The systemic risk measured by the DebtRank and marginal expected shortfall models shows monotonicity with the bank type, while the risk measured by △CoVaR does not. The systemic risk of different types of banks changes dynamically in different years. The systemic risk measured by DebtRank is positively correlated with both size and centrality, and Cai demonstrates that, of the three models, DebtRank performs best at measuring the systemic importance of banks from the perspectives of size and interconnectedness. The results of his paper provide empirical evidence for the systemic importance of banks.

The issue’s second paper has the formidable title “Internet financial risk assessment in China based on a particle swarm optimization–analytic hierarchy process and fuzzy comprehensive evaluation”. The paper’s authors, Zeng Li, Wee-Yeap Lau and Elya Nabila Abdul Bahri, detail how the combination of finance and the internet has expanded the boundaries of and markets for financial services. Due to the problems of virtualization and information technology dependence, the risk management of internet finance is more complicated than that of traditional finance. Based on the identification of China’s internet financial risk, Li et al’s study establishes a comprehensive evaluation system to index internet financial risk. They use a particle swarm optimization–analytic hierarchy process and a comprehensive fuzzy evaluation of an expert questionnaire to show that these evaluation methods are highly applicable to internet financial risk assessment. In addition, their findings show that credit and technical risk are very important, and that internet finance risk is high in China. The authors therefore recommend that the supervision of internet financial risk in China be strengthened and that more attention be paid to the internet finance credit regulation and technical risk processes. While this application is local, its implications are global and this paper offers further evidence that risk always front-runs regulation.

“Value-at-risk and the global financial crisis” by Manh Ha Tran and Mai Ngoc Tran is the third paper in this issue. Using daily data for seven large international banks, Tran and Tran examine the forecasting ability of bank value-at-risk (VaR) estimates around the 2007–9 global financial crisis (GFC) period. They find that the internal VaR estimates from their sample of banks are very inaccurate and systematically overstate the VaR during the precrisis and postcrisis periods, with a mixed performance during the GFC. Some banks inflated their VaRs, while others experienced excessive VaR exceptions and clustering. VaR estimates based on simple models of the generalized autoregressive conditional heteroscedasticity (GARCH) type easily outperform banks’ internal VaR estimates. The VaR estimated via GARCH under the Student t distribution (ie, GARCH-t) captures the extreme losses reasonably well. Tran and Tran attribute the poor VaR estimates at banks to the banks’ inappropriate choice of internal VaR models. Although this result is unsurprising, it is rather disturbing. At the time of the GFC, banks’ risk modeling was underfunded and neglected. Perhaps things are different now.

The issue’s final paper, “Does the asymmetric exponential power distribution improve systemic risk measurement?” by Shu Wu, Huiqiong Chen and Helong Li, looks at the choice of distribution to model financial returns. As many have noted before, the return distributions involved in the measurement of systemic risk using parametric modeling suffer from fat tails, asymmetric kurtosis and asymmetric tails. This immediately rules out symmetric distributions such as the normal or Student t . Wu et al advocate the use of asymmetric exponential power distribution (AEPD), which can potentially avoid overfitting and underfitting problems because, among other things, it can be reduced to a Gaussian distribution or a generalized error distribution. They implement a parametric estimation for the systemic risk measure conditional value-at-risk (CoVaR) of Huang and Uryasev, and they compare the goodness-of- fit and backtesting performance of the AEPD with other commonly used distributions (the normal, Student t and skewed-t distributions). Based on data from the Chinese banking sector from 2008 to 2019, their empirical results show that AEPD has the best goodness-of-fit. Moreover, it is the only distribution that provides a validated estimation for CoVaR. Readers will find the clear detailing of steps used in the calculations very helpful in carrying out validation exercises themselves.