Integration Between the Chinese Agricultural and U.S
Energy Market: Realized Volatility Transmission
Evidence from a Multivariate HAR Model
1 | P a g e
DongYihai Peng Peter Louis Sanusie
1310754 1086247
Authors’ Note:
This paper’s findings, interpretations and conclusions are entirely written for the purpose of major assignment as
part of the FNCE40004 Numerical Techniques in Finance assessment at The University Of Melbourne. To the
best of the authors’ knowledge, whose name written above, there has not been a submitted paper with similar
materials and empirical methods except for the papers which are referenced in this paper. The author claims no
liability and is not responsible in any way to any person or entity who plans to use this paper for business or other
purposes. The authors would also like to thank Dr. Jonathan Dark for his guidance and teaching of this subject.
FNCE40003 Numerical Techniques in Finance Major Assignment
2 | P a g e
Abstract
This paper implements the extension of HAR model: Multivariate Orthogonalized HAR
model, which is able to decompose short, mid and long-term spill-over effects in a multivariate
setting. Moreover, the second moment of realized volatility can be examined to find evidence
of market integration by applying DCC / CCC GARCH model to the error term of Multivariate
Orthogonalized HAR model. We examine the volatility transmission effect using high
frequency data of Chinese soybean, Chinese soybean oil, Chinese corn, Chinese palm oil, US
WTI and US Ethanol futures after the global financial crisis, starting from 1 January, 2010 to
30 April, 2022. This is mainly motivated by scarcity of studies using Multivariate HAR model
and studies on spill-over effect between energy and agricultural market, particularly in
international setting. We find a statistically significant weekly and monthly spill-over effect
from WTI to Chinese soybean but weak weekly reversal effect and a two-way direction in the
Granger Causality test. Moreover, we find a statistically significant coefficient for daily and
monthly spill-over effect from Chinese corn to Ethanol with a relatively low magnitude but no
reversal effect and a one-way direction in the Granger Causality test from Chinese corn to
Ethanol. Examining the residuals of Multivariate Orthogonalized HAR Model within CCC
GARCH framework, we find no clear or limited evidence of market integration between the
two market over time.
Keywords:
Volatility Transmission
Multivariate HAR model
Dynamics Conditional Correlation GARCH
Constant Conditional Correlation GARCH
Energy Futures Market
Agricultural Futures Market
FNCE40003 Numerical Techniques in Finance Major Assignment
3 | P a g e
Table of Contents
1. INTRODUCTION................................................................................................................ 4
2. LITERATURE REVIEW ................................................................................................... 8
3. METHODOLOGY ............................................................................................................ 10
3.1. REALIZED VOLATILITY ................................................................................................ 10
3.2. VHAR MODEL .............................................................................................................. 11
3.2.1 HAR Component ..................................................................................................... 11
3.2.2 Transmission Component ....................................................................................... 12
3.2.3 Variance and Time-varying Correlation of Realized Volatility using DCC
GARCH ............................................................................................................................ 12
3.2.4 Estimating Parameter Using Maximum Likelihood Function ............................. 14
3.3. DIEBOLD YILMAZ INDEX .............................................................................................. 15
4. DATA .................................................................................................................................. 16
4.1 CHOOSING FUTURES CONTRACT DATA ........................................................................ 16
4.2 DESCRIPTIVE DATA FOR REALIZED VOLATILITY AND TREATMENTS FOR REALIZED
VOLATILITY ......................................................................................................................... 17
4.3 TESTS FOR THE HYPERGEOMETRIC DECAY PROCESS OF REALIZED VOLATILITY ..... 18
5. RESULTS ........................................................................................................................... 20
5.1 THE SIMPLE HAR MODEL REGRESSIONS .................................................................... 20
5.2 THE BIVARIATE MHAR MODEL REGRESSIONS THE WTI-CHINESE AGRICULTURE
SPILL-OVER EFFECTS ........................................................................................................... 22
6. CONCLUSION .................................................................................................................. 31
7. REFERENCES ................................................................................................................... 32
8. APPENDIX ......................................................................................................................... 34
FNCE40003 Numerical Techniques in Finance Major Assignment
4 | P a g e
1. Introduction
With the recent wave of globalization coming from a digitalized world, there has been
an increase in the interconnectedness across financial markets in the world. The so-called
‘Capital Market Integration’ may likely to derive from the activities of international investors
and multinational companies, economic integration by means of deregulation and linearization,
and market contagion (Heimonen 2002, p. 415).
This phenomenon has attracted many research interest and numerous studies. Papers
such as Fratzscher (2002), Li and Majerowska (2008), Yi and Tan (2009), Nor (2012), Berger
and Pozzi (2013), Muharam, Wisnu and Arfinto (2019), and numerous other papers have
studied the stock market integration across different regions
1
. Most of these studies implement
the Multivariate Generalized Autoregressive Conditional Heteroskedasticity (Multivariate
GARCH) or the extension of it to examine the volatility transmission across markets.
However, one possible drawback of many of these studies is that most employ a sample
of daily returns or even lower frequency returns which are known to be subject to noise (Soucek
and Todorova 2013, p.586). Therefore, this paper will follow Soucek and Todorova (2013) in
applying a multivariate extension of Heterogenous Autoregressive (HAR) model as it utilizes
realized volatility from intraday returns. The univariate version of HAR model is universally
recognized in the econometric literature due to its practicality and ability to be expanded with
external variables. In the context of multivariate extension of HAR model, Soucek and
Todorova (2013, p. 586) states that its primary advantage is the capability to split the
transmission effect over daily, weekly, and monthly horizon which is not achievable using
multivariate GARCH models.
1
See Heimonen (2002) for a review on studies of stock market integration before 2000
FNCE40003 Numerical Techniques in Finance Major Assignment
5 | P a g e
This paper then examines the volatility transmission of the U.S crude oil and
agricultural products in China using the multivariate extension of HAR model. Oil and
agriculture are arguably crucial factors in the macroeconomy, and their volatility will have a
major influence in the macroeconomic dynamics. Although it is commonly established that the
price of agriculture is influenced by energy (oil) prices by means of transportation and
production cost, the increase demand of ethanol as an alternative fuel in agriculture productions
have created a view of stronger interconnectedness between the two markets (Beckmann &
Czudaj 2014, p.119). Ethanol as an alternative fuel is produced from corn as its main ingredient,
thus there might be spill-overs from corn to ethanol market
2
. Additionally, market fundamental
and arbitrage-based economic theory suggests interrelation between oil, ethanol and corn
(sugar) prices (De Gorter and Just 2008).
Studying volatility transmission between agricultural and energy market might shed
light on several matters in finance. Wu et al (2011) discover that increase in the volatility of
crop prices alter the hedging and investment decisions by increasing the cost to manage price
risks for farmers. Moreover, the co-movement between may worsen portfolio diversification
as there has been an increase of inclusion for agricultural commodities in many portfolios
(Gardebroek and Hernandez 2013, p.119). Gardebroek and Hernandez (2013, p.119) also argue
that volatility spill-over within agricultural commodities via substitution in supply and demand
may be problematic for importing countries in which their consumer are dependent on food in
a large scale. Lastly, an increase of agricultural price volatility impact the price stabilisation
policies’ effectiveness and design at the macro-level (Byrne et al. 2011).
In particular, we study the volatility transmission between agricultural market in China
and energy market in U.S. China ranked first for oil consumption and import in 2020, reaching
approximately 13 million barrels per day
3
. This should provide us with a great stage to conduct
our empirical analysis.
2
U.S Department of Energy, ‘Ethanol Production’, https://afdc.energy.gov/fuels/ethanol_production.html
3
Statista, ‘Leading global crude oil importers in 2020’, https://www.statista.com/statistics/240600/global-oil-
importers-by-region-2011/