主管:中国科学院
主办:中国优选法统筹法与经济数学研究会
   中国科学院科技战略咨询研究院

中国管理科学 ›› 2025, Vol. 33 ›› Issue (3): 80-92.doi: 10.16381/j.cnki.issn1003-207x.2022.0634cstr: 32146.14/j.cnki.issn1003-207x.2022.0634

• • 上一篇    下一篇

基于MIDAS-SVQR的供应链金融质押物风险价值测度新方法

汪刘凯, 张小波, 王未卿(), 刘澄   

  1. 北京科技大学经济管理学院,北京 100089
  • 收稿日期:2022-03-29 修回日期:2022-08-10 出版日期:2025-03-25 发布日期:2025-04-07
  • 作者简介:王未卿(1973-),女(汉族),天津人,北京科技大学经济管理学院,金融工程系主任兼书记,副教授,研究方向:金融科技与大数据分析、供应链金融、金融风险管理,E-mail: wangwq@manage.ustb.edu.cn.
  • 基金资助:
    国家自然科学基金项目(72301025);教育部人文社会科学研究规划基金项目(20YJA630024);中国博士后科学基金项目(2021M700380);北京市社会科学基金项目(23GLB022);中央高校基本业务研究基金项目(FRF-DF-20-11)

MIDAS-SVQR: A Novel Model for Measuring VaR of Supply Chain Finance Pledge

Liukai Wang, Xiaobo Zhang, Weiqing Wang(), Cheng Liu   

  1. The School of Economics and Management,University of Science and Technology Beijing,Beijing 100089,China
  • Received:2022-03-29 Revised:2022-08-10 Online:2025-03-25 Published:2025-04-07

摘要:

存货质押作为供应链金融的典型融资方式,质押物价值波动是供应链金融面临的主要风险之一,因此,如何测度质押物价格波动风险是学界和业界关注的焦点。VaR作为Basel协议主推的风险度量工具,已被学界和业界广泛使用。然而,关于VaR测度的现有方法存在:收益分布误设、非线性关系刻画不准确和混频数据信息提取不充分等潜在挑战,因此,本文提出了一种测度供应链金融质押物VaR的新方法:MIDAS-SVQR。一方面,该方法基于分位数框架下利用核函数捕获非线性关系以直接输出分位数,而无需分布假设;同时,利用MIDAS处理混频数据,提升其利用混频数据信息的能力。此外,本文基于二次规划详细给出了MIDAS-SVQR的求解过程。最后,本文选取钢铁、铜等六种典型质押物为研究对象,选择GARCH类和QR类等模型作为基准模型,并基于Kupiec检验等三种回测方法来评价模型准确性。结果表明:MIDAS-SVQR在所有样本的三种回测检验下表现最优。此外,分位数回归类模型总体表现明显优于GARCH类模型。因此,本文提出的MIDAS-SVQR新方法既有效度量了供应链金融质押物的风险价值,也为供应链金融风险管理提供了新技术支持。

关键词: 供应链金融, VaR, MIDAS-SVQR, 混频数据, 支持向量分位数回归

Abstract:

Pledged inventory is one of the typical financing modes of supply chain finance (SCF), and the fluctuation of pledge value is the main risk faced by SCF. Therefore, how to measure the risk of pledge value fluctuation is the focus of SCF's risk management. as many previous works have shown, Value at Risk (VaR), which is mainly promoted by the Basel accord, is widely used in academic and industry for risk measures. However, the conventional VaR measures approaches have three challenges: 1) the true distribution of the data is not known, so the distribution assumption is prone to mistakes; 2) it is difficult to accurately describe the nonlinear relationship between variables; 3) the mixed frequency data is not fully utilized. To address the above issues, the combination of mixed data sampling (MIDAS) and support vector quantile regression (SVQR), namely MIDAS-SVQR model, are first applied to improve the performance of pledge’s VaR measure. The novel approach uses kernel functions to deal with nonlinear relationships and directly outputs quantiles without any distribution assumptions; meanwhile, it uses MIDAS to process the mixed frequency data to increase the ability of the model to extract the information from the mixed frequency data. To illustrate the efficacy of our method, empirical studies on six representative pledges include steel, copper, lead, zinc, and tin. The data is collected from Wind (https://www.wind.com.cn/) and covers the period from Jan 1, 2007 to August 31, 2021. Then, the proposed model is compared with the classical model (GARCH), quantile regression (QR), SVQR and MIDAS-QR in terms of Kupiec test, conditional converage test and VaR duration test. The empirical results are promising and show that our method (with the highest average P value of the three backtests across all samples) outperforms the others. Moreover, it is found that the quantile regression models generally perform significantly better than the GARCH models. In future, reversed (un)restricted MIDAS can be incorporated into SVQR to enable model to use more mixed frequency data. To this end, this is an interesting topic and we leave it for future research.

Key words: supply chain finance, Value at Risk, MIDAS-SVQR, mixed-frequency data, support vector quantile regression

中图分类号: