國際指標原油價差之研究

The Study of the Spread of International Index Crude Oil Price

林淑娟(Shu Chuan Lin)

指導教授 ： 張四立 ； 廖惠珠 ； 林茂文

國立臺北大學/公共事務學院/自然資源與環境管理研究所/博士(2016年)

摘要

本研究以一系列計量檢測工具分析探討國際主要指標原油價差趨勢，利用結構轉變點分析方法探索及驗證其價差序列之結構性轉變時點，透過分量單根檢定分析主要指標原油價差在高低分量回到均值(mean-reventing)的能力。期能有效掌握國際主要指標原油價差之變化，以做為石油公司在市場結構丕變之際，在生產、行銷及輸儲規劃上作出妥適因應及機動應變之參考。
本研究分析Brent-WTI、Brent-Dubai與WTI-Dubai三種國際主要指標原油價差之單根檢定與因果關係後，在結構轉變點探討方面，先利用Bai & Perron、Chow檢定、移動式Chow檢定等三種結構性轉變檢定方法，分別對各類原油價差日、週與月資料進行檢測。分析結果顯示，各方法所檢測出的結構性轉變時點具有相當的一致性。之後，再以時間數列離群值偵測方法進行分析，而獲致相似之結果。
另外，本研究也進行分量單根檢定。結果發現，三種原油價差日、週與月資料之高低分量價差，雖區域性或全球化現象呈現情形有所差異，但QKS檢定值在整個資料期間均有共整合現象，表示三種價差雖在部份期間呈現脫勾現象，但整體代表仍會回復均值。其中，Brent-WTI與Brent-Dubai價差均顯示愈低分量，其半衰期愈低，亦即愈穩定；WTI-Dubai價差分析結果則顯示愈高分量，其半衰期愈低，亦即愈穩定。而三種價差中的半衰期以Brent-Dubai最小，亦即相對穩定，其次為Brent-WTI、WTI-Dubai。這些結果隱含上述原油價差結構轉變現象屬暫時性現象。只要資料期間夠長，原油價差終將回復均值。
最後，以月資料的時間變化參數(TVP)檢視WTI原油價格對Brent-WTI價差的影響，發現隨時間的演進，其影響在2010年年底之後呈現明顯變大的情形。將上述月資料的結構轉變時點以虛擬變數方式納入時間數列轉換模式後，比較納入前之MAPE，結果顯示考量結構性轉變有助於配適度之提升。綜述，分析國際主要指標原油價差，運用結構性轉變與分量單根檢定方法了解資料特性，有助於提升預測模型之配適績效與預測準確度。

關鍵字

國際原油價差 ； 結構轉變分析 ； 分量單根檢定 ； 時間變化參數 ； 時間數列轉換函數

並列摘要

This study aims to explore and realize the movement and trend of international crude oil benchmark price and their price spreads by using the structural change analysis to detect and examine the structural change point of crude oil price spread, and the quantile unit root test to analyze the mean-reverting trait for higher and lower crude oil price spreads. The analyzed results could provide as a reference for oil companies to properly plan the strategies of production, marketing, and storage and pipeline construction when the structure change occurs in the oil market. In doing so, three international crude oil benchmark price spread, i.e., Brent-WTI, Brent-Dubai, and WTI-Dubai, were examined by using unit root test and Granger causality test. And then, several structural change analysis methods, i.e., Bai & Perron test, Chow test, moving Chow test, and time series outliner detecting analysis, were applied to detect the structural change point for three crude oil price spreads with daily, weekly, and monthly time series data. The analyzed results show that the structural change points detected with different methods are consistent and robust.
In term of quantile unit root test, although the localized or globalized phenomenon for daily, weekly, monthly data of three crude oil price spreads with higher or lower quantiles are different, QKS test statistics indicated the co-integration phenomenon for the whole time series period. This shows that three crude oil price spreads deviate from mean within some time series period but revert to the mean within whole time series period. Among them, Brent-WTI and Brent-Dubai price spreads with lower quantile while WTI-Dubai price spread with higher quantile have lower half-life, that means that more stable characteristics. Else, the half-life of Brent-Dubai price spread are the smallest, while the half-life of Brent-WTI and WTI-Dubai price spreads are larger. Furthermore, time varying parameter (TVP) module was adopted to examine the impact of WTI on Brent-WTI price spread in time series transfer function, the result shows the impact extent of WTI on Brent-WTI price spread are getting larger as time elapses. Moreover, dummy variable represented structural change point of time series was included into model to evaluate the MAPE performance, the result indicated that goodness of fit of the model increased after considering structure change in the model. In summary, this study suggested that the methods of structural change and quantile unit root test should be used to understand the characteristics of time series before analyzing the movement and trend of international crude oil benchmark price. This helps to enhance the goodness of fit and the forecast performance of the forecasting model.

並列關鍵字

International Crude Oil Price Spread ； Structural Change Analysis ； Quantile Unit Root Test ； Time Varying Parameter ； Time Series Transfer Function

參考文獻

Chinese Petroleum Corporation (CPC), Taiwan. (2005) The impact analysis on motor gasoline consumption in Taiwan, IAEE Conference.

彭竣永(2010)，通貨膨脹持續性:分量單根檢定之應用，國立政治大學經濟學系碩士論文。

李怡慧(2008)，「國際原油價格波動主因之研究」，淡江大學，產業經濟學系博士論文。

Abraham, B. & Box, G. E. P. (1979) Bayesian analysis of some outlier problems in Time Series, Biometrika, 66. 229-236.

Adelman, M. A. (1984) International Oil Agreements, Energy Journal, 5, 1-9.