Realized Volatility Forecasting of Agricultural Commodity Futures Using Long Memory and Regime Switching

• Author: Fengping Tian,Langnan Chen,Ke Yang
• Date: 01 July 2017
• DOI: http://doi.org/10.1002/for.2443
• Published date: 01 July 2017

Realized Volatility Forecasting of Agricultural Commodity Futures

Using Long Memory and Regime Switching

FENGPING TIAN,

1

KE YANG

2,3

AND LANGNAN CHEN

4

*

1

International School of Business and Finance, Sun Yat-sen University, Guangzhou, Guangdong China

2

School of Economics and Commerce, South China University of Technology, Guangzhou, Guangdong

China

3

College of Economics and Management, South China Agricultural University,Guangzhou, Guangdong

China

4

Lingnan College, Sun Yat-sen University, Guangzhou, Guangdong China

ABSTRACT

We investigate the dynamic properties of the realized volatility of five agricultural commodity futures by employing

the high-frequency data from Chinese markets and find that the realized volatility exhibits both long memory and

regime switching. To capture these properties simultaneously, we utilize a Markov switching autoregressive fraction-

ally integrated moving average (MS-ARFIMA) model to forecast the realized volatility by combining the long

memory process with regime switching component, and compare its forecast performances with the competing models

at various horizons. The full-sample estimation results show that the dynamics of the realized volatility of agricultural

commodity futures are characterized by two levels of long memory: one associated with the low-volatility regime and

the other with the high-volatility regime, and the probability to stay in the low-volatility regime is higher than that in the

high-volatility regime. The out-of-sample volatility forecast results show that the combination of long memory with

switching regimes improves the performance of realized volatility forecast, and the proposed model represents a

superior out-of-sample realized volatility forecast to the competing models. Copyright © 2016 John Wiley & Sons, Ltd.

key words realized volatility; forecast; agricultural commodity futures; long memory; regime switching

INTRODUCTION

Traditionally, the fractionally integrated generalized autoregressive conditional heteroscedasticity (GARCH) class models

have been utilized to forecast the volatility of agricultural commodity futures in the literature. As long memory is found in

most volatility series for agricultural commodity futures, the introduction of fractional integration will improve the model’s

forecast performance (see Crato and Ray, 2000; Jin and Frechette, 2004; Baillie and Kapetanios, 2007; Coakley et al., 2008;

Hyun-Joung, 2008; Sephton, 2009; Chang et al., 2012). However, due to the unobservable nature of the volatility, these

studies treat the volatility of agricultural commodity futures as a latent process in the GARCH class models.

The availability of high-frequency data with more intraday trading information has enabled the application of

advanced volatility proxies such as the realized volatility based on the sum of squared intraday returns (Andersen

and Bollerslev, 1998), which allows us to treat the latent volatility as an observed variable and model the volatility

series directly, rather than treated it as a latent process in the GARCH class models. Many properties of the realized

volatility have been well documented in the recent literature. One of the most relevant properties is that the realized

volatility exhibits high persistence, as evidenced in Andersen et al. (2003) and Corsi (2009). For this reason, linear

autoregressive fractionally integrated moving average (ARFIMA) models are used to capture this property. A

flexible strategy to model the serial dependencies for the realized volatility is proposed by Barndorff-Nielsen

(2002) through a superposition of the ARMA(1,1) processes, whereas Corsi (2009) develops a heterogeneous

autoregressive (HAR) model given by a combination of volatilities measured over different time horizons.

Nevertheless, financial time series are subject to occasional structural breaks due to important events such as

financial crisis, major changes in market sentiments, generation of speculative bubbles and regime switches in

monetary policies (Stock and Watson, 1996; Pesaran and Timmermann, 2002). Recently, some studies (e.g. Liu

and Maheu, 2008) have shown the evidence of structural breaks in realized volatility, and found that the volatility

forecast without taking into account the breaks will result in overestimation of the long memory parameter and poor

forecast performance (Rapach and Strauss, 2008; Choi et al., 2010). Other studies have been doubtful about the

adequacy of long memory models for volatility series, as long memory can be easily affected by the structural

breaks. In particular, it is suggested that various types of structural changes can lead to a strong persistence in

the autocorrelation function, and hence generate ‘spurious’long memory. Granger and Hyung (2004) present the

evidence that the spurious long memory can be detected from a short memory process with level shifts. Diebold

*Correspondence to: Langnan Chen, Lingnan College, Sun Yat-sen University, Guangzhou, Guangdong, China.

E-mail: lnscln@mail.sysu.edu.cn

Journal of Forecasting,J. Forecast. 36, 421–430 (2017)

Published online 27 September 2016 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/for.2443

Copyright © 2016 John Wiley & Sons, Ltd.