原油價格建模與預測

Modeling and Forecasting Crude Oil Prices

許玉雪(Esher Hsu) ； 謝沛珊(Pei-Shan Hsieh)

《統計與資訊評論》 20卷 (2020/12) Pp. 39-68

摘要

本研究旨在透過時間序列分析建立原油價格之預測模式。研究方法包含實證分析及模擬分析，首先利用單變量ARCH／GARCH、Markov Switching Model（MS）模型以及多變量ARMA模型（VARMA）進行實證比較，挑選出較適模型後，再利用Monte Carlo法進行模擬分析，最後整合實證分析及模擬分析結果，找出最佳預測模型做為原油價格的預測模式。實證分析採用2000年1月至2017年12月WTI、Dubai、Brent原油價格月平均資料。實證結果發現MS-ARI及VARI模型的配適結果較佳。進一步進行Monte Carlo模擬比較MS-ARI及VARI的預測能力，模擬結果發現MS-ARI有較佳的預測能力，然而VARI模式的預測能力也不差。整合實證分析與模擬分析結果發現MS（2）-ARI（1,1）是原油價格的較佳預測模型，因此建議MS（2）-ARI（1,1）為原油價格的預測模式。

關鍵字

原油價格 ； 時間序列模型 ； GARCH ； Markov Switching Model ； VARMA

並列摘要

This study aims to carry out a better model for price forecasting of crude oil based on time series analysis. The methodology of this study includes empirical analysis and simulation analysis. Three models, namely ARCH/GARCH, Markov Switching Model (MS), and Vector ARMA (VARMA) are used for fitting crude oil prices in empirical study. The selected better time series models based on the results of the empirical analysis are further compared by Monte Carlo Simulation to find the best model for crude oil price forecasting. Monthly average crude oil prices of WTI, Dubai, and Brent from January 2000 to December 2017 are used for the empirical study. Empirical results show that the MSARI and VARI models have better performance on model fitting. The MS-ARI and the VARI are thus selected for further simulation study to compare their prediction performance. Simulation results indicate that the MS-ARI model has better performance, but the VARI model also performs well. Study results show that the MS(2)-ARI(1,1) has better performance in both empirical study and simulation study. Therefore, it is suggested for the prediction model of crude oil prices.

並列關鍵字

Crude oil prices ； Time series model ； GARCH ； Markov Switching Model ； VARMA