期刊
ENERGY ECONOMICS
卷 60, 期 -, 页码 35-46出版社
ELSEVIER
DOI: 10.1016/j.eneco.2016.09.020
关键词
Bayesian forecasting; Dynamic model averaging; DMA; Dynamic model selection; DMS; Forecasting oil price; Oil price; Predicting oil price; Spot oil price; Time-varying parameters
类别
资金
- Polish National Science Centre [DEC-2015/19/N/HS4/00205]
This paper is aimed on the analysis of monthly spot oil prices (WTI) between 1986 and 2015. The methodology is based on Dynamic Model Averaging (DMA) and Dynamic Model Selection (DMS) framework The important feature of DMA method is an allowance for both time-varying coefficients and large state space model (i.e., the set of oil price determinants can change in time). Within this framework it was explicitly shown how the significance of oil price determinants vary in time. These determinants itself were chosen with respect to some previous studies. Contrary to the currently reported DMA applications in some other fields, no significant evidence was found that DMA is superior over, for example, ARIMA model. However, DMA could also not been rejected as a significantly worse model due to certain statistical tests. The performed DMA analysis was checked for robustness on various model parameters and for certain computational issues. It was found, for example, that in the context of the 2008 oil price peak exchange rates and stock markets were important oil price drivers, whereas oil production or oil import were just minor determinants. Some role of the change in inventories was found, but not greater than the one in 1991. The role of China's economy as an oil price driver in 2008 was found to be relatively smaller than in other time periods. Also, the robustness of these findings was discussed. (C) 2016 Elsevier B.V. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据