基于Copula-CVaR-EVT方法的供应链金融质物组合优化

为缓释当下供应链金融业务单一质物价格剧烈波动诱发的贷款集中度风险,异于股票、债券等金融资产组合基于短期风险预测优化框架,提出一类更具普适性的基于蒙特卡罗模拟法的质物组合长期风险预测方法,克服现有长期风险预测中视为基准的时间平方根法则缺陷;比对银行采取积极和保守投资策略,建立基于均值CVaR质物组合优化框架,引入改进均值方差优化框架进行对比分析.为准确测度质物组合长期CVaR,建立ARMA-EGARCH-EVT族模型以及多元tCopula模型,刻画现货质物收益率呈现出的自相关性、"尖峰厚尾"以及波动集聚性等典型事实特征以及质物间的非线性相关结构;从模型层面和研究对象层面进行敏感性分析以验证模型的稳健性以及结论的可靠性.实证结果显示:长期风险预测视角下均值CVaR框架较改进的均值方差模型更具优势,为风险限额管理下的商业银行提供一种组合质物风险管理的新框架和新模式.

Inventory portfolio optimization in supply chain finance: A Copula-CVaR-EVT approach

This paper proposes a mean-CVaR portfolio optimization framework for both conservative and aggressive investment strategies based on long-term risk prediction, which is different from financial assets such as stocks, bonds portfolio optimization framework based on the short-term risk prediction, so as to mitigate concentration risk due to sharp fluctuations of price of single inventory in supply chain finance. The long-term risk prediction based on Monte Carlo simulation of the inventory portfolio is proposed, and it is more practical than square root rule, which overcomes the shortcoming of the square root rule which heavily depends on the independent normal distribution. In methodology, AR(1)-EGARCH(1,1)-EVT model is set up to better depict the characteristics of the autocorrelation, heteroskedasticity, leptokurtosis and fat-tails of the marginal distribution, furthermore, the multivariate t-Copula function is introduced to model the dependency structure of individual pledged inventory. The empirical results show that, the mean-CVaR optimization framework outperforms the improved mean-variance from the perspective of long-term risk prediction, which are robust to the choice of risk window, confidence level, simulation times and sample size. In summary, this paper provides a new framework for managing the risk of portfolio in inventory financing practice for banks constrained by risk limitation.