A separate reduced‐form volatility forecasting model for nonferrous metal market: Evidence from copper and aluminum

• Author: Haibo Liu,Yaoqi Guo,Xuehong Zhu,Hongwei Zhang
• Date: 01 November 2018
• Published date: 01 November 2018
• DOI: http://doi.org/10.1002/for.2523

RESEARCH ARTICLE

A separate reduced‐form volatility forecasting model for

nonferrous metal market: Evidence from copper and

aluminum

Hongwei Zhang

1,2

| Xuehong Zhu

1,2

| Yaoqi Guo

2,3

| Haibo Liu

4

1

School of Business, Central South

University, Changsha, China

2

Institute of Metal Resources Strategy,

Central South University, Changsha,

China

3

School of Mathematics and Statistics,

Central South University, Changsha,

China

4

Konka Group Co. Ltd, Shenzhen, China

Correspondence

Yaoqi Guo, School of Mathematics and

Statistics, Central South University,

Changsha 410083, China.

Email: guoyaoqi@csu.edu.cn

Funding information

National Natural Science Foundation of

China, Grant/Award Numbers: 71633006,

71403298 and 71701176; National Social

Science Foundation of China, Grant/

Award Number: 13 & ZD169

Abstract

This article extends the HAR‐CJN model proposed by Andersen, Bollerslev, and

Huang (Journal of Econometrics, 2011, 160, 176–189) and explores the role of

overnight information and leverage effects in improving volatility forecasting.

To explore the interaction between different components of daily volatility, this

paper attempts to separately model the dynamics of continuous variation, the

discontinuous jump, and the overnight return variance by including leverage

effects. The findings show that lagged continuous and discontinuous jump var-

iations generate significant impacts on future continuous segments, discontin-

uous jump segments, and the overnight return variance. Furthermore, in

addition to the usual leverage effects, additional leverage effects with respect

to overnight returns are found to play a significant role in volatility forecasting.

Finally, out‐of‐sample forecasts are investigated; the results show that the new

HAR‐CJN model can describe and predict daily volatility more accurately than

other HAR models.

KEYWORDS

HAR‐CJN model, leverage effects, metalfutures, overnight information,volatility forecasting

1|INTRODUCTION

As the Chinese economy continues its rapid development,

demand for base metals is burgeoning. Metals continue to

play a significant role in both industrial manufacturing

and world economic activities. In recent years, increasing

speculation in emerging economies has resulted in more

uncertainty and volatility in metal markets, making vola-

tility forecasting in metal markets an attractive and valu-

able subject for financial traders and manufacturers

alike. Volatility forecasting can contribute to investment

decisions about portfolio allocation, value‐at‐risk manage-

ment (Brooks & Persand, 2003), and manufacturers'

industrial production. Consequently, accurate volatility

forecasting in the metal futures market is a crucial issue

for both financial traders and policymakers.

Early studies on volatility modeling and forecasting in

the metal market primarily focused on generalized

autoregressive conditional heteroskedasticity (GARCH)‐

type models (Behmiri & Manera, 2015; Bentes, 2015;

Hammoudeh, Malik, & McAleer, 2011; Hammoudeh &

Yuan, 2008; Kristjanpoller & Hernández, 2017;

Kristjanpoller & Minutolo, 2015; McKenzie, Mitchell,

Brooks, & Faff, 2001). More recent literature suggests that

high‐frequency data can significantly improve the predic-

tion accuracy of future volatility (Gong, He, Li, & Zhu,

2014; Wang & Wang, 2016; Wen, Gong, & Cai, 2016).

Moreover, the decomposition between continuous and

discontinuous jump components can contribute to acquir-

ing more accurate forecasts. Furthermore, the possible

leverage effects (if the market exhibits this particular fea-

ture) may obtain more‐accurate forecasts (Duan, Chen,

Received: 13 December 2017 Accepted: 10 February 2018

DOI: 10.1002/for.2523

754 Copyright © 2018 John Wiley & Sons, Ltd. Journal of Forecasting. 2018;37:754–766.wileyonlinelibrary.com/journal/for