管理评论 ›› 2021, Vol. 33 ›› Issue (7): 16-28.

• 经济与金融管理 • 上一篇    

北美市场原油和天然气联动性研究——基于贝叶斯DCC-GARCH和LSTAR模型的实证分析

柴建1, 林婕1, 梁婷2   

  1. 1. 西安电子科技大学经济与管理学院, 西安 710126;
    2. 湖南大学工商管理学院, 长沙 410082
  • 收稿日期:2018-05-11 发布日期:2021-08-02
  • 通讯作者: 林婕(通讯作者),西安电子科技大学经济与管理学院硕士研究生;梁婷,湖南大学工商管理学院博士研究生
  • 作者简介:柴建,西安电子科技大学经济与管理学院教授,博士生导师,博士。
  • 基金资助:
    国家自然科学基金面上项目(71473155);陕西省青年科技新星项目(2016KJXX-14);西安电子科技大学基本科研业务费项目(JB160603)。

A Study of Linkage between Crude Oil and Natural Gas in North American Market: Based on the Empirical Analysis of Bayesian DCC-GARCH Model and LSTAR Model

Chai Jian1, Lin Jie1, Liang Ting2   

  1. 1. School of Economy and Management, Xidian University, Xi'an 710126;
    2. School of Business, Hunan University, Changsha 410082
  • Received:2018-05-11 Published:2021-08-02

摘要: 油气互为替代品的特征使得二者间表现出一定关联性。北美作为全球最大的油气消费市场,其油气价格的变动及联动将影响全球的能源发展格局及投资策略。文章首先基于贝叶斯DCC-GARCH (动态条件相关多变量广义自回归条件异方差)模型对北美市场原油和天然气收益率之间的整体相关性进行研究。在油气收益率整体相关性较不显著的情况下,通过STL分解研究了二者在趋势、季节和随机项上的动态关联性特征。随后进一步运用LSTAR (Logistic平滑转换自回归)模型分析了原油和天然气趋势走向上的动态关联性的非线性特征。研究结果表明,北美市场油气收益率的动态关联性始终为正向,关联区间为[0137,0216],中位数为0181,较CCC-GARCH模型(常数条件自相关广义自回归条件异方差模型)的结果更能反映油气间联动的时变性特征,但其关联区间变动幅度较小,不存在显著的相关性。STL分解结果显示,油气收益率在趋势走向上存在显著的关联性,LSTAR模型较AR模型能够更好地刻画油气趋势联动性的波动特征,具有更高的拟合程度。油气趋势关联性在数值2026处发生平滑转换,从低区制向高区制转换的速度很快。该研究结果对投资者提高收益、生产者控制成本、国家相关部门制定相应的能源策略等具有较强的指导意义。

关键词: 原油, 天然气, 联动, 趋势, 非线性

Abstract: The characteristic of oil and gas as substitutes for each other makes the two show a certain correlation. The changes and linkage of oil and gas prices in the world's largest oil and gas consumer market will affect the global energy development pattern and investment strategy. Firstly, the paper studies the overall correlation between crude oil and natural gas yields in North America based on the Bayesian DCC-GARCH (Dynamic Conditional Correlation GARCH) model. In the case that the overall correlation between the two was less significant, the dynamic correlation characteristics of the two in trend, season and remainder are studied by STL decomposition. And then we analyze the nonlinear characteristics of correlation between oil and gas in trend through the LSTAR (Logistis Smooth Transformation Autoregressive) model. The results show the dynamic correlation between crude oil and natural gas yields in North American market is always positive, with a correlation range of[0.137,0.216] and a median of 0.181, the result of which could better reflect the time-varying characteristics of oil and gas linkage compared with the CCC-GARCH model (Constant Conditional Correlation GARCH). The correlation range of oil and gas is small, showing that there is no significant correlation between them. The results of STL decomposition show that there is a significant correlation between oil and gas yields in trend. And, LSTAR model could better characterize the correlation fluctuation and have a higher degree of fitting than AR model. The correlation between oil and gas in trend has a smooth transition at 2.026, and the transform speed from low zone to high zone is very fast. The results of the study have guiding significance for investors to increase their income, producers to control their costs, and the relevant government departments to formulate corresponding energy strategies.

Key words: crude oil, natural gas, linkage, trend, nonlinear