基于QRNN+GARCH方法的供应链金融多期价格风险测度及防范

Evaluating and Preventing Multi-period Price Risks in Supply Chain Finance via QRNN+GARCH Method

在供应链金融中,存在多期价格风险,如何有效测度和防范,成为学界与业界共同关注的重要问题.考虑到供应链金融业务的多期性,质押物收益的非线性、非对称、波动聚焦性等特征,充分发挥神经网络分位数回归（QRNN）模型在非线性与非对称方面的处理能力,结合GARCH模型在波动聚集性准确刻画方面的优势,提出了一个QRNN＋GARCH的建模框架,给出了供应链金融多期价格VaR风险测度;其次,为对比新测度方法与GARCH模型等传统方法的优劣,基于似然比检验与平均相对误差,对VaR风险测度效果进行评价,给出了相应的评价方案;再次,为确定与设计合理的多期质押率,达到有效防范风险之目的,给出了风险不可控比率和效率损失率两个指标,评价质押率的有效性。最后,选取现货铝为对象,实证研究其价格波动行为,结果表明：第一,在多期价格风险测度方面,QRNN＋GARCH方法明显优于GARCH模型,表现为准确性更高,更具效率和稳健性;第二,在防范风险方面,QRNN＋GARCH方法所确定的多期动态质押率,能够更好地降低效率损失。

In supply chain finance, there often exists multi-period price risks. How to evaluate and prevent the risks is an important question for academics and practitioners. Firstly, considering the natures in multi-period supply chain finance business, such as nonlinearity, asymmetry, and volatility clustering, we develop a QRNN＋GARCH procedure and advance a VaR measure for evaluating multiperiod price risks in supply chain finance. The procedure takes the advantages of the quantile regression neural network（QRNN） model＇s ability in processing the asymmetry and nonlinearity and the GARCH model＇s superiority in describing volatility clustering. Secondly, we provide an approach to evaluate the performance of our new method and GARCH model, using the likelihood ratio test and the average relative error. Thirdly, to design a rational impawn rate and the effectiveness of preventing risk, we construct two useful indices： efficiency loss rate and uncontrollable risk ratio, which are applied to check the efficacy of impawn rate. Finally, an empirical analysis is conducted on the spot aluminum.The empirical results show that the QRNN＋GARCH method is preferable than the GARCH model in evaluating multi-period price risks since the former can obtain more accurate, efficient and robust results. In addition, when it comes to risk prevention, the efficiency loss can be significantly reduced by the dynamic impawn rate determined via the QRNN＋GARCH approach.