Figure 2.2 Maize Production in Indonesia Source: BPS-Statistics Indonesia
2.3.3 Soybean
In Indonesia, soybeans are commonly used for producing tofu and tempeh. As a secondary crop, soybeans are less attractive for farmers compared to other
food crops, such as rice and maize, which are more profitable. This has stimulated land-use competition and as long as the relative price of soybeans is lower than the
other two commodities farmers will have low incentive to grow soybeans on their land Agustine 2014. As shown in Figure 4 below, the domestic soybean
production has decreased from 1.01 million tonnes in 2000 to 954 thousand tonnes in 2014. Moreover, the national demand for soybeans, accounting for 2.23 million
tonnes, exceeds domestic production encouraging the dependency on imported soybeans.
5 10
15 20
25
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
m il
li o
n to
nnes
Maize Production
Figure 2.3 Soybean Production in Indonesia Source: BPS-Statistics Indonesia
2.3.4 Sugar
Similarly, production of sugar in Indonesia has failed to meet national sugar demands. According to Saleh Husin, the Indonesian Minister of Industry, domestic
demand for sugar is about 5.7 million tonnes, consisting of 2.8 million tonnes of white sugar for household consumption and 2.9 million tonnes of refined sugar for
industry. As shown in Figure 5 below, such demand significantly exceeds domestic production which only accounted for 2.58 million tonnes in 2014 Hasibuan 2015.
Low sugar production is mainly caused by inefficient production processes of 100 year old sugar factories leading to limited production capacity and low quality sugar
Hasibuan 2015.
0.2 0.4
0.6 0.8
1 1.2
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
Soybean Production
Figure 2.4 Sugar Production in Indonesia Source: BPS-Statistics Indonesia
2.3.5 Wheat
In terms of wheat, Indonesia has become one of the largest wheat importers in the world. The domestic demand for this commodity is high, but farmers are
reluctant to plant wheat on their land due to the inappropriate characteristics of the land and climate in Indonesia for growing wheat. As a consequence, Indonesia
imports wheat in large quantities, at approximately 5.4 million tonnes in 2014, from several countries such as Australia, Canada, USA, Russia, Ukraine, Kazakhstan,
India, Pakistan, Brazil and Argentina Jaramaya 2015.
2.3.6 Crude Palm Oil
In contrast, Indonesia is known as one of the largest crude palm oil exporters in the world dominating the world palm oil production. Palm oil can be used to
produce biodiesel, RBD palm oil refined, bleached and deodorized crude palm oil including cooking oil and margarine, as well as oleochemicals. In 2014, domestic
demand for crude palm oil is about 8.8 million tonnes, consisting of 2.5 million tonnes for biodiesel industries, 4.2 million tonnes for RBD palm oil industries and
0.7 million tonnes for oleochemical industries Fahriyadi and Laoli 2015. As shown in Figure 6 below, domestic production of crude palm oil largely surpasses
domestic consumption. This low domestic demand enables Indonesia to export cooking oil products to several countries including China and India.
0.5 1
1.5 2
2.5 3
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
tho us
an d
to nnes
Sugar Production
Figure 2.5 Crude Palm Oil Production in Indonesia Source: BPS-Statistics Indonesia
Overall, shocks in oil price can be transmitted onto food crop commodities through production costs and biofuel demands. This means that fluctuations in oil
prices are expected to influence food prices, posing risks to national food security. Moreover, rising oil prices encourage the use of biofuels as an alternative energy
source. In Indonesia, biofuels are mainly produced from palm oil, corn and sugar cane. These three commodities are not only used for feedstocks, but also for food.
This then leads to lower food supply thereby contributing to higher food prices. These factors combined induce a significant rise in food commodity prices posing
threats to food security in Indonesia.
5 10
15 20
25 30
35
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
m il
li o
n to
nnes
Crude Palm Oil
3 LITERATURE REVIEW
An increasing number of studies have investigated the relationship between crude oil and food prices. These studies show a divergence in findings indicating
that the effect of crude oil price volatilities on food commodity prices is not universal. This has motivated researchers worldwide to further investigate such
correlations. The majority of previous studies have highlighted the existence of volatility spillover from crude oil prices to food commodity prices, an effect that is
becoming stronger as biofuel production increases. Such research concurs with crude oil price volatilities being transmitted onto food commodities prices through
two essential elements, namely production inputs transportation and electricity costs, fertiliser and pesticide prices and biofuel production agricultural
commodities as raw materials.
Research investigating volatility spillovers from crude oil onto the food market through production inputs has focused on agricultural commodities used for
staple food and found a strong relationship between them. Research by Baffes 2007 was based on 35 non-energy commodities traded globally, such as food rice,
maize, soybeans, wheat, and sugar, beverages cocoa, coffee, and tea, raw materials cotton, rubber, and timber, fertilisers, and metals silver, aluminum,
copper, nickel, zinc. This author used annual price data from 1960 to 2005 and analysed data using an Ordinary Least Squares OLS regression method. Results
indicate that linkages between crude oil prices and food commodities are stronger compared to other commodities. Furthermore, research results of Balcombe 2011
also suggest a positive correlation. This author’s findings, acquired using the time varying and panel approach, were based on smaller samples covering 19
agricultural commodities, such as rice, soybeans, maize, wheat, meat, cheese, cocoa,
and sugar from 1957 to 2009. A more recent study by Obadi and Korček 2014 investigated the causality, long-run and short-run linkages between price of crude
oil and food products, such as palm oil, wheat, corn, sugar, rice and barley. Data used in this study were monthly commodity prices over the period January 1975 to
September 2013 and by adopting a Granger causality and a VECM model, a long- run relationship between oil and food prices was identified. In addition, in the long-
run, causality is transmitted from oil prices to all food commodity prices, except for barley which has a two-way causality. Similarly, in the short-run, the direction of
causality runs from oil prices to each of the food commodities, except for sugar. These findings, therefore, suggest that the price of agricultural food commodities
used as sources of food react significantly to oil price fluctuations as oil is a primary input in energy intensive farmin
g Obadi and Korček 2014. In order to investigate the impact of the biofuel boom on volatility
transmission between oil and other commodity prices, previous studies have concentrated on agricultural products being used as raw materials for biofuels and
divided analyses into two time periods, namely prior to and after the 2008 global food crisis. Serra 2011 used weekly price data of international crude oil and the
weekly price of ethanol and sugar from July 2000 to November 2009 to investigate volatility transmission between crude oil, ethanol and sugar prices. The methods
used in this study were the semiparametric GARCH and parametric MGARCH. Results indicate that in the long run, crude oil, ethanol and sugar price levels are
connected by an equilibrium parity, suggesting a strong volatility linkage between them. Research by Ji and Fan 2012 uses daily log returns data covering various
crops soybean, wheat, corn, and sugar and metals gold, silver, aluminum between 7 July 2006 and 30 June 2010. By adopting an EGARCH model, these
authors suggest that the nexus between crude oil prices and agricultural and metal commodity prices has become stronger since the 2008 crisis. The work of Bakhat
and Würzburg 2013, derived from an ECM model, is based on three categories of samples which are food commodities used to produce biofuel, such as soybean oil,
sunflower oil, palm oil and sugar, other food commodities which cannot be converted into biofuels, such as rice, wheat, and beef and agricultural commodities
which are not edible such as rubber, coffee and wool. By analysing monthly data for the period January 2000 to April 2011, these authors found that increasing the
use of biofuels induces stronger linkages between crude oil and food commodities, especially for commodities used as raw materials for biofuel production. Similarly,
recent research by Tadesse, Algieri, Kalkuhl and Braun 2014 which used monthly and annual data of oil and food prices, such as wheat, corn and soybeans, from 1986
to 2009, indicates that in recent years there is an emerging link between food, energy and financial markets. Thus, such results indicate an indirect transmission
between oil price volatility and farming commodities through increased demand for biofuel.
Other researchers state that results of investigations into volatility transmission from oil onto food commodity prices are not statistically significant at
all times. A study by Du, Yu and Hayes 2011 shows that linkages between oil, maize and wheat only exist after 2006. They analysed the weekly data of these
commodities by following a stochastic volatility model and a Bayesian econometric method for estimating the parameters of the model. Similar results were obtained
by Nazlioglu, Erdem and Soytas 2013 who adopted the Hafner and Herwartz test approach and divided the data into two time periods of observation: before the food
crisis 01 January 1986-31 December 2005 and after the food crisis 01 January 2006-21 March 2011. The data used in the study were daily price data of
agricultural commodities, such as wheat, maize, soybeans and sugar. The results indicated that during the pre-crisis period there was no volatility spillover between
crude oil and agricultural commodity prices. In contrast, there was significant volatility transmission during the post-crisis period. Hence, these findings show that
the existence of volatility spillovers may depend on a particular time period marked by the biofuel boom phenomenon after the crisis Nazlioglu, Erdem and Soytas
2013.
Other studies suggest an insignificant effect of volatility spillover between the oil and biofuel markets. Zhang, Lohr, Escalante and Wetzstein 2010 used a
VECM model to show that there is no direct long-run relationship between fuel and agricultural commodity prices asserting that if there were any direct short-run
relationships, then they would be finite, except for sugar because it can affect agricultural product prices through rising biofuel production. These results were
obtained by analysing monthly price data for energy commodities, such as gasoline, ethanol and oil, and farming products, such as rice, maize, soybeans, sugar and
wheat, covering the period March 1989 through July 2008. Similar results were reported by Kaltalioglu and Soytas 2011, who expanded the sample to include a
greater range of commodities. These authors used the monthly price data of oil,
food products fruits, vegetables, meat, poultry, fish, grocery food and non- alcoholic beverages and agricultural raw commodities timber, cotton, wool,
rubber, and hides between January 1980 and April 2008. By adopting a Granger causality in variance approach, they found that there is no volatility spillover from
energy market to food and agricultural raw commodity markets. Gardebroek and Hernandez 2013 used weekly price data of crude oil, ethanol and corn from
September 1997 to October 2011. Based on the results obtained from an MGARCH model, there is a higher correlation between ethanol and corn markets, particularly
after 2006. However, these authors did not find evidence that energy price volatilities stimulate corn prices in the US. A recent study by Abdelradi and Serra
2015 adopted parametric and semiparametric methods suggests similar results. They used weekly price data for the period 06 November 2008 through 14 June
2012. Results indicate that the biofuel industry in Europe does not trigger an increase in food prices. These contradictory findings have motivated researchers to
further analyse the relationships between crude oil price volatilities and food commodity prices in order to shed light on the existence of volatility transmission
in different countries.
4 DATA AND METHODOLOGY
This section presents data and explores the econometric methods used to investigate the relationship between oil prices in the world market and food
commodity prices in the Indonesian market from 2002 to 2015. To measure such a linkage this study adopts the VECM model and the Granger Causality test.
4.1 Data
To investigate volatility spillovers between energy and food markets, this study uses a series of individual prices for commodities rather than price indices in
order to circumvent aggregation bias and the weighting rule when making them into indices Arezki, Hadri, Loungani and Rao 2013. The data set includes a monthly
average price series for world crude oil, world food commodities and Indonesian food prices from January 2002 to August 2015, consisting of 164 observations.
The West Texas Intermediate WTI crude oil price is used to represent world oil prices. Food commodities selected in this study are rice, maize, wheat, soybeans,
sugar and cooking oil. Both crude oil and world food commodity prices are compiled from the World Bank database. Domestic food commodity prices are
obtained from the Indonesian Ministry of Trade and converted into US. All commodity prices are in real rates 2010=100 and expressed in natural logarithms.
Details of the commodity data are presented in Table 4.1 below.
Table 4.1 Data of Commodities Commodities Variable
Description Unit
Source
Crude Oil CO
Crude oil, US, West Texas Intermediate WTI 40` API
US Dollars per barrel USbbl
World Bank World Rice
WRC Rice Thailand, 5 broken,
white rice US Dollars per metric
tonne USmt World Bank
World Maize WMZ
Maize US, no. 2, yellow US Dollars per metric
tonne USmt World Bank
World Wheat WWHT
Wheat US, no. 1, hard red winter
US Dollars per metric tonne USmt
World Bank World Soybean
WSOY Soybean US
US Dollars per metric tonne USmt
World Bank World Sugar
WSGR Sugar US
US Dollars per kilogram USkg
World Bank World Palm Oil
WPLM Palm oil Malaysia, 5
bulk US Dollars per metric
tonne USmt World Bank
Domestic Rice LRC
Medium quality rice US Dollars per metric
tonne USmt Ministry of Trade
Domestic Maize
LMZ Local maize
US Dollars per metric tonne USmt
Ministry of Trade Domestic
Wheat LWHT
Segitiga biru US Dollars per metric
tonne USmt Ministry of Trade
Domestic Soybean
LSOY Local soybean
US Dollars per metric tonne USmt
Ministry of Trade Domestic Sugar
LSGR Local sugar
US Dollars per kilogram USkg
Ministry of Trade Domestic
Cooking Oil LCOOK
Yellow palm oil US Dollars per metric
tonne USmt Ministry of Trade
Domestic Petroleum
PO Domestic Petroleum
US Dollars per barrel USbbl
Ministry of Energy and Mineral
Resources
4.2 Methodology
This study uses four basic models to analyse linkages between price of crude oil and the price of the following domestic food commodities: rice, maize, wheat,
soybeans, sugar and cooking oil. The first model is used to test whether fluctuations in crude oil prices can directly affect domestic food prices, as represented by the
equation:
�
�
= + �
�
+
�
where �
�
is the price of the domestic food commodity � at time � ; �
�
is the price of crude oil at time
� ; and
�
is the error terms. The second model tests whether crude oil prices influence change in domestic
food prices through domestic petroleum prices, given that domestic petroleum production uses crude oil as a primary input, as expressed by the following
equations:
�
= + �
�
+
�
�
�
= +
�
+
�
where
�
is the price of the domestic petroleum at time � ; �
�
is the price of crude oil at time
� ; �
�
is the price of the domestic food commodity � at time
� ;
�
and
�
are the error terms.
The third model is used to investigate whether volatilities of crude oil prices are transmitted through to food commodities in the world market, expressed as:
�
= + �
�
+
�
�
�
= � + �
�
+
�
where
�
is the price of the world food commodity � at time � ; �
�
is the price of crude oil at time
� ; �
�
is the price of the domestic food commodity � at time � ;
�
and
�
are the error terms. To investigate existence of a stable linear steady-state link between these
variables, this study follows a Unit Root test, Co-Integration test, Vector Error Correction Model and Granger Causality test.
4.2.1 Unit Root Test
Economic time series data is generally characterised by stochastic properties and strong trend behaviour. According to Hatanaka 1996, these properties can
violate stationarity conditions which require a constant mean, constant variance and constant covariance over time. Granger and Newbold 1974 state that if non-
stationary time series data is used in a regression model it can generate a spurious regression result leading to false existence of a significant economic relationship
from unrelated variables. Therefore, before doing a time series analysis it is important to test for the stationarity properties of the time series data.
The Unit Root test can be used to determine the stationarity characteristics of time series data by detecting the presence of a unit root. This study adopts a
commonly used non-stationarity test known as the Augmented Dickey-Fuller ADF test. The null hypothesis of this test is that the series has a unit root, meaning
that it is not stationary Obadi and Korček 2014. This hypothesis can be rejected if the value of the t-statistic is lower than the ADF critical values, which are
substantially higher in absolute terms than those of standard regression.
4.2.2 Cointegration Test
It is believed that two or more price series are cointegrated if these prices move simultaneously in the long run. Engle and Granger 1987 point out that if a
linear combination of two or more variables, which are integrated of the same order � � , generate a stationary series then the series are cointegrated. This means that a
long run equilibrium relationship between economic variables may exist even though they deviate in the short run.
This study uses the Johansen Cointegration test to analyse the existence of a long run relationship between variables. Before carrying out the test it is important
to determine the order of integration of variables and ascertain that these variables share the same order of integration. The null hypothesis of this test is that there is
no cointegration relationship between variables.
4.2.3 Vector Error Correction Model VECM
The existence of cointegration between variables indicates a long term equilibrium relationship between them. However, this does not specify short run
effects of the cointegrated variables. Thus, VECM can be applied to determine the short run dynamics of each variable in a frame that binds such dynamics to their
long run equilibrium relationships Zhang, Lohr, Escalante and Wetzstein 2009.
VECM is a restricted form of traditional VAR. Such a restriction has to be applied due to the existence of cointegration between the non-stationer variables in
the level and it then incorporates information from the cointegration residuals into the model. The specification of VECM restricts the long-term relationships of
endogenous variables so that they can converge to their cointegration relationships, but still allows the dynamics of the short run Chintia 2013. Hence, by using
VECM it is possible to analyse the long run and short run causality between oil and each food price. The following equations are the regression form of VECM:
Δ
�
= ∑ Δ
�− �−
=
+ ∑ Δ
�− �−
=
+ � �
�−
+
�
Δ
�
= ∑ Δ
�− �−
=
+ ∑ Δ
�− �−
=
+ � �
�−
+
�
where and denote dependent or independent variables; , , and
are parameters which will be estimated; � and � are coefficients of error
correction variables; �
�−
and �
�−
denote lagged residuals from the regression of cointegration;
�
and
�
are error terms. If
� and � are equal to
zero, then this indicates that the equations are a traditional VAR model in the first difference. Inversely, if
� and � are significantly different from zero, then Δ
�
is affected by the deviation from the long run equilibrium, from the previous period,
meaning that
�
has an error correction effect. This indicates that VECM is the appropriate estimation method because it is a VAR model in the first-differenced
form which includes cointegration residual vectors Chen and Patel 1998.
4.2.4 Granger Causality Test
The Granger Causality test is used to analyse whether two economic variables have at least one way of causality. This method estimates how much current values
of can be explained by past values of . According to Chen and Patel 1998, is granger caused by if can help to predict and the coefficients of the lagged
s are statistically significant. There are two possibilities of causality, one-way and two-way causation.
This study adopts the traditional pairwise Granger Causality test to examine the causality relationship between crude oil prices and food commodity prices. This
method requires the test to be done at the levels of the data C. The following equation is the pairwise Granger Causality as a bivariate autoregressive model:
�
= � + ∑
�− �
=
+ ∑
�− �
=
+
�
�
= � + ∑
�− �
=
+ ∑
�− �
=
+
�
where � denotes the maximum number of lagged variables incorporated into
the model; , , and are the coefficients of the model representing each
lagged variable contribution to the predicted value of
�
and
�
;
� , � are
constants and
�
,
�
are error terms.