Volatility forecasting of crude oil market: A new hybrid method

• Date: 01 December 2018
• Published date: 01 December 2018
• Author: Jin‐Liang Zhang,Yue‐Jun Zhang
• DOI: http://doi.org/10.1002/for.2502

RESEARCH ARTICLE

Volatility forecasting of crude oil market: A new hybrid

method

Yue‐Jun Zhang

1,2

| Jin‐Liang Zhang

3

1

Business School, Hunan University,

Changsha, PR China

2

Center for Resource and Environmental

Management, Hunan University,

Changsha, PR China

3

School of Economics and Management,

North China Electric Power University,

Beijing, PR China

Correspondence

Jin‐Liang Zhang, School of Economics and

Management, North China Electric Power

University, Beijing 102206, PR China.

Email: zhangjinliang1213@163.com

Funding information

National NaturalScience Foundation of

China, Grant/Award Numbers: 71774054,

71101011, 71322103 and 71431008; National

Special SupportProgram for High‐Level

Personnel fromthe Central Government of

China; Changjia ng Scholars Program of the

Ministry of Education of China;Hunan

Youth Talent Program; Beijing Philosophy

and Social SciencePlanning Project, Grant/

Award Number: 2014BJ0264

Abstract

Given the complex characteristics of crude oil price volatility, a new hybrid

forecasting method based on the hidden Markov, exponential generalized

autoregressive conditional heteroskedasticity, and least squares support vector

machine models is proposed, and the forecasting performance of the new

method is compared with that of well‐recognized generalized autoregressive

conditional heteroskedasticity class and other related forecasting methods.

The results indicate that the new hybrid forecasting method can significantly

improve forecasting accuracy of crude oil price volatility. Furthermore, the

new method has been demonstrated to be more accurate for the forecast of

crude oil price volatility particularly in a longer time horizon.

KEYWORDS

crude oil, EGARCH, HMM,LSSVM, volatility forecasting

1|INTRODUCTION

Volatility forecasting of crude oil prices has been proven

to be an important input into multifactorial decision

making processes, including macroeconomic policy

making, financial risk assessment such as value‐at‐risk

(VaR) calculations, options pricing and portfolio manage-

ment strategies (Sadorsky, 2006; Xu & Ouenniche, 2012).

Considering the importance of crude oil in economic

growth over recent years, price volatility forecasting

has received increasing attention from governments,

investors, analysts and academics, and abundant quanti-

tative forecasting methods have been developed, which

may be divided into three categories: (i) time series

volatility models; (ii) implied volatility models; and (iii)

hybrid models.

J. L. Zhang, Zhang, and Zhang (2015) presented a

relatively systematic review of existing literature

concerning typical models used to forecast crude oil prices

and their volatility. In order to accurately forecast the

volatility of crude oil prices, it is essential to scientifically

identify the appropriate volatility models. In the past,

generalized autoregressive conditional heteroskedasticity

(GARCH) class models have often been employed to

describe the volatility of crude oil prices. These models

perform well in effectively modeling the time‐varying var-

iance and clustering features of the conditional

variance of crude oil price returns. However, traditional

GARCH models (such as the commonly used GARCH(1,

1) model) require that all parameters should be positive

to ensure the positive variance of crude oil market

returns. Given that the volatility of crude oil prices often

Received: 30 November 2015 Revised: 27 July 2017 Accepted: 16 November 2017

DOI: 10.1002/for.2502

Journal of Forecasting. 2018;37:781–789. Copyright © 2017 John Wiley & Sons, Ltd.wileyonlinelibrary.com/journal/for 781