International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
Modeling Price Volatility of Nigerian Crude Oil Markets Using
GARCH Model: 1987-2017
Deebom, Zorle Dum & Isaac Didi Essi
Department of Applied Mathematics/Statistics,
Rivers state University, Port Harcourt
dumzorle@yahoo.com
Abstract
Volatility and the risk-return trade off of crude oil or crude oil market participation is
essential to National Investment, decision making, marketing, and the determination of the
financial strength of Nations among other things. Therefore, this research study was targeted
at modeling price volatility and the risk-return related to crude oil export in Nigerian crude
oil market using the first order asymmetric and symmetric univariate Generalized
Autoregressive Conditional Heteroscedasticity (GARCH) family model in three distributional
assumptions namely, Normal, student’s-t and generalized error distribution. To achieve this
target, three objectives with three research questions and two hypotheses were raised for the
study. The data for the study was extracted from the Central Bank of Nigeria online statistical
database starting from January, 1987 to June, and 2017. The results from the statistical
analysis reveal that the markets were optimistic of their investment and other trade related
activities. Sequel to that, there were high probabilities of gains than losses. Although, the
variables use in these markets were extremely volatiles and shows evidence there exists
positive risk first-rated meaning that investments or investors deserved rewards for holding
risky assets. In estimation, first order symmetric GARCH model (GARCH, (1,1) in student’s-t
error assumption gave a better fit than the first order Asymmetric GARCH model (EGARCH
(1,1)) in Normal error distributional assumptions. However, the selected models were
subjected to several diagnostic test such as ARCH effect test, test for serial correlation and
QQ-plot in order to validate their fitness which was confirmed to be appropriate. And
recommendations were made to the Government to look for new ways to diversify the
economy from total dependence on oil to non-crude oil such as agriculture, manufacturing
and mining sector. For investors or marketers in this markets, they were advice to be mindful
in trading in a highly volatile period especially when there is evidence of high standard
deviation in the descriptive statistic of the return series and in modeling volatility of price
return of certain micro/ macro-economic variable the leverage effect of such variable should
be properly estimated using asymmetric GARCH model.
Key Words: Modeling, Volatility, Crude Oil Price, GARCH Models, Markets
1.0 Introduction
Financial time series data such as stock prices, exchange rates, inflation rates, crude oil prices
etc. are some of the variables that often exhibit the characteristics of clustering. A period
otherwise refers to as volatility clustering whereby prices show wide swings within an
extended time and it will later show relatively calmness. This is not only applicable to
variables such as stock prices, exchange rate, inflations etc. but they also applicable to almost
all micro economic variables. For instance, all the indicators and determinant of employment
and production, consumption, investment in raising productive capacity and how much a
country imports and exports (John, 2003) also suffered the same fate. They suffered sudden
fluctuation and this continual fluctuation affects so many things thereby contributing to the
IIARD – International Institute of Academic Research and Development
Page 23
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
increase in price volatility and revenue profile of these products. And these are some of the
causes of economic shocks widely experience in the world. According to Agenor et al (2000)
the macroeconomic effects of macro econometric variable and trade shocks arising from price
volatility have a great very significant effect on developing countries.
These shocks are major sources of aggregate economic volatility and they have large impact
on private and public savings because of their economic effects (Agenor et al., 2000). They
are also associated with global business cycles and it manifest in the form of sharp volatility
in foreign exchange earnings of primary producing economics as in the case of Nigeria. Such
development usually results in macroeconomic instability, in sufficient allocation of
resources, recessions and low output growth.
According to Gujarati (2009) the awareness of volatility is of crucial importance in many
areas. For example, considering it sudden sharp changes in prices investors and traders alike
cannot know the appropriate time to invest and when not to as a result of instability in
world’s prices. This does not guarantee safer investment especially now that crude oil market
and other financial market like stock and foreign exchange markets are more dependent on
each other than ever before. For traders in these markets or decision markers, volatility in it
entirety may not be bad, but its variability may not be good enough because this makes
financial planning cumbersome.
This is also applicable to the importers, exporters and traders in foreign exchange markets,
this variability in the exchange rates may account for excessive losses or profits. According
to Gujarati (2009) investors in the stock market are obviously interested in the volatility of
stock price, for high volatility could mean huge losses or gains and hence greater uncertainty.
In volatile markets such as the crude oil markets, it will be difficult for companies to raise
capital in the crude oil markets.
In crude oil market, when there is a sharp fall in the international oil price and which may
lead to corresponding consequent decline in financial receipts as case in the early 1980’s
when the economy can no longer meet it international financial commitments. These make
nations to be tangle with situations that could become a big challenge. So the questions are
how do we model financial time series that exhibit such characteristics behavior? For
example, how we model time series of crude oil prices? A characteristic exhibited by crude
oil prices such that in its level form it could be liken to random walks or called it stochastic
process. That is, a situation that shows they are not stationary. Conversely, in the first
difference form, they become stationary as it is in the case of other micro economic variable
like GDP series. The usual traditional regression tools have proved their limitation in the
modeling of high-frequency (weekly, daily of intra-daily) data (shamiri et al, 2009).
Shamiri et al., (2009), further suggested that assuming the only the mean response could be
changing with covariates while the variance remain constant within time varying interval, it
will often revealed to be an unrealistic assumption in practice. This fact is particularly clear in
special time series data where there exist clusters of volatility such that it is visually detected.
Although, in the past view decade there have been several forms of different propositions on
how to model such characteristics exhibited by price in the form of heteroscedasticity.
According to Shamiri et al (2009), among the models that have be proven to be most
successful are the Auto-regressive conditional heteroscedasticity (ARCH) family model
originally invented by Engle (1982) and the models of stochastic variance (SV) pioneered by
IIARD – International Institute of Academic Research and Development
Page 24
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
Taylor. Engel (1982) argues that an adequate volatility model is the one that sufficiently
model heteroscadasticity in the disturbance term and also captures the stylized fact inherent
in stock return series such as volatility clustering, Autoregressive Conditional
Heteroscadasticity (ARCH) effect and asymmetry.
This is one of the reason why we model variance in financial series data as well make
forecast, which is very important in many areas where option price is to be examine, value at
risk apply and portfolio consideration. Therefore, it becomes necessary to model out ofsample forecasting ability as a natural model selection conditions for volatility models.
Although, there are numbers of variance forecasting research carried out in this area, some
researchers used squared daily returns as a substitute for ex-post variance but this has been
proven by Anderson et al (1998) to be an unbiased and above all a noisy estimator. While
some other literatures that review competing variance models has been neglected due to other
necessary conditions needed for effective volatility model. Meanwhile, very little work has
been done comparing different error distribution assumptions, with the remarkable exceptions
as opined by Shamiri et al (2009).
However, none of these studies has actually focused on modeling asymmetric GARCH
models forecast with respect to their error distributions. Majority of the previous research
studies in this area are often done on the symmetric GARCH model, especially on stock
returns, exchange rates etc while this studies focus on both symmetric and asymmetric
volatility as well as their various symmetric and asymmetric error distribution assumptions on
crude oil export price.
Most likely, ARCH models usually do not fully capture the thick tail characteristic as
exhibited in the form of fat-tailed distribution. Of a truth, Kurtosis of most price return series
is greater than three, invariably means that extreme values are observed more often for such
variable than normal distribution. So long as Kurtosis of the return series is a well-established
condition, the situation is peculiar to symmetry of such distribution. Many researchers seem
to ignore the significant of this while other researchers, the likes of Simkowits and Beedles,
(1980) and Kon (1984) have drawn our attention to the characteristics exhibited by heavy tail
of a distribution (Shamiri et al, 2009), whereas Shamiri et al (2009) have shown that fat-tailed
distribution is a requirement for modeling daily returns of East Asian Equity market as cited
in Shamiri et al, (2009). Hence, this study will among other things add to the existing
literature in several important ways. It is on this ground that the study model volatility of
crude oil prices in us/per barrel from January, 1987 – June, 2017 in order proffer solutions to
these problems.
2.0 Literature Review
Literature examining the GARCH modeling otherwise called the generalized autoregressive
conditional heteroskedasticity model is a very complex concept that captured and measured
volatility characteristics exhibited by most micro as well as macro-economic variables. These
micro and macro-economic variables could be export prices, exchange rate, Gross Domestic
product (GDP) etc. This model measure unequal variance and it effect on other micro or
macro-economic indicator in the economy.
According to Atoi (2014), the first break-through in modeling variable that exhibit such
characteristics was championed by Engle, (1982). Engle (1982) demonstrated that conditional
heteroskedasticity can be modeled using conditional variance of the (Atoi, 2014).
IIARD – International Institute of Academic Research and Development
Page 25
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
Engle (1982) demonstrated that conditional heteroskedasticity can be modeled using an autoregressive conditional variance of the disturbance term with the linear (combination of the
square disturbance in the recent past second past see Atoi (2014) Having realized the
potentials of an Autoregressive conditional heteroskedasticity (ARCH) model several studies
have using it in modeling model financial series.
However, when using the ARCH model in determining the optimal lag length of variables are
very cumbersome. Therefore, often time users encounter problems of over parameterization.
Thus, Rydberg (2000) argued that since large lag values are required in ARCH model
therefore there is the need for additional parameters.
Sequel to that, and many other Lapses and little minor challenges encountered in the ARCH
model, (Bollerslve et al, 1986) independently proposed an extension to ARCH model which
was refers to as Autoregressive Moving Average (ARMA). This was done with view to
achieving parsimony. And this eventually lead to the development of the model called the
generalized Autoregressive conditional heteroskedacity (GARCH), which model conditional
variance as a function of its lagged value of the disturbance term of linear regression model.
Although, GARCH model have been proven to be useful in capturing symmetric effect of
volatility but the model is bedeviled with some limitations such as relatively non-negative
constraints imposed on the parameters to be estimated. Therefore, this study among other
things investigates as well test GARCH family model performance in modeling price
volatility.
Sequel to the above , there is the need to review other numerous studies carried out in this
area, the method they employed, the countries and the result obtained. There are several
studies carried out in this area but the general observation from these studies is that the results
have been characterized with mixed feeling, depending on several factors including sample
period, methodology adopted, estimation technique, measure of volatility adopted and the
countries under consideration (either developed or developing).
For example, Tatyana (2010) studied the dynamics of oil prices (Brent and WTI crude oil
markets) and their volatilities by Linking four GARCH related models namely; GARCH
(1,1), GJR – GARCH (1,1), EGARCH (1,1) and APARACH (1,1). The findings of this study
showed that oil shocks have permanent impact and there exist asymmetric consequence on
the volatility of the markets under consideration.
Also, Abduchakeem et al (2016) in analyzing oil price – macroeconomic volatility in Nigeria
using GARCH model and its variants (GARCH-M, EGARCH and TGARCH) using daily,
monthly and quarterly data. From the findings the result reveal that: all the macroeconomic
variables considered (real gross domestic product, interest rate, exchange rate and oil price)
are highly volatile; the asymmetric models (TGARCH-M) suggested that oil price is a major
source of economic volatility in Nigeria.
In like manner, Narayan et al (2007) modeled the volatility of daily oil prices using
exponential generalized Autoregressive conditional heteroskedasticity (EGARCH) model.
The results also reveal that asymmetric effects exist, persistent, and is permanent in the oil
prices series. Also, Olowe (2009) investigated weekly oil price volatility of all countries
average spot price using EGARCH (1,1) within January 3, 1997 – March 6, 2009. The result
shows that the oil price return series has high persistence of volatility, volatility clustering
and asymmetric characteristics.
IIARD – International Institute of Academic Research and Development
Page 26
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
Olugbenga et al (2017) study the impact of oil price volatility on investment decision making
in marginal fields development in Nigeria. The study also investigated the relationship
between oil price volatility and marginal field investment analysis in Nigeria. The marginal
field’s crude oil production was used as a replacement of investment analyze. A monthly data
from October, 2015 – April 2016 was considered. The GARCH model, Johansen cointegration and Granger causality tests were used in estimating the results. However, the
result showed a significant positive relationship between oil price volatility and crude oil
production (P < 0.05).
Omisakin (2008) no an analysis of oil prices stocks on the Nigeria economy using an annual
data on seven key macro-economic variables, from 1979-2005, vector Autoregressive model
was used in estimating variables and it was pointed out that oil price shocks contribute to
variability in the economic price.
The concept and overview of price volatility according to olugbenga et al (2017), suggested
that the econometric terms, volatility is defined as the rate at which the price of a security
increases or decreases in a given set of returns. Volatility is measured by returns. It is
measured by calculating the standard of deviation of either daily or monthly or the yearly
returns of stocks price over a given period of time. It shows the extent to which the price of a
certain products may increase or decrease. If the prices of a certain products fluctuate rapidly
in a short time period, it indicates volatility is on the increase. If the prices of a certain
products fluctuate slowly in a longer interval of time, it clearly shows that volatility is low.
Although, Atoi (2014) suggested that an increase or decrease in the value of stock return
tends to have a corresponding effect on the economy, mostly through the money market; an
increase in product prices can motivate investment and increases the demand for credit,
which eventually leads to increase interest rates in the overall economy as supported by
(Spiro, 1990).
Hence, there is the need to develop an appropriate volatility model to captured variations in
product price returns which is of significant policy importance to econometricians and
economic managers alike. Most especially, reliable volatility model for crude oil export
prices returns that will guide traders, investor, Government agencies etc in their risk control
management decisions and portfolio selection.
Modeling can be seen as a process of simplifying system used to simulate some aspects of the
real economy (see John, 2002). In this context, the real economy could be refers to as price
volatility. The characteristics behavior of price to violate the normality assumption
(Homoskedasticity) other wises refers to as heteroskedasticity that lead to the introduction of
the concept of modeling volatility. Heteroskedasticity, according to Olugbenga et al, (2017)
is one of the key problems that require attention when performing time series analysis on
crude oil price due to uncertainty in the movement of oil prices. The sudden up and down in
the movement of crude oil export price is referred to as price volatility. And this can be
model econometrically using the residual conditional variance of the regression equation
involving crude oil export price as the dependent variable.
Although, according to Ederington et al (2011) Models for oil prices can be classify into three
main categories: Structural models, reduced form or hybrid and econometric model. The
structural models are basically used in capturing the interaction between primary or supply
and demand conditions and factors influencing supply. It tends to focus on longer timehorizons which may include macro-type models used for forecasting. The Reduced form or
IIARD – International Institute of Academic Research and Development
Page 27
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
hybrid models on the other hand leverage on the hypotheses about the reduced form to
examine the stochastic behavior of oil prices, whereas the Econometric models hypothesize
specific types of time series behavior in the conditional first and second moments of crude oil
price series. Reduced form or hybrid and econometric has the tendency to focus on short-term
dynamic behavior of crude oil prices.
The significant measurement typically missing from present models of crude oil prices but
which has become an issue of great importance to oil market observers and participants is the
role that speculators play in the market and the implications of such role on crude oil sale.
However, this study will be constrain to used econometric models form as it has the tendency
to model residual of both mean and conditional variance of the dependent variable (Crude
oil export price). In order to achieve that, Engle (1982) suggested that Autoregressive
conditional heteroskedasticity (ARCH) model is the alternative model to the standard
deviation, method. The ARCH model is the only time series model that can best and
sufficiently model heteroscadasticity in the disturbance term of a linear regression equation
involving crude oil price as the dependent variable and also captures the stylized fact
incorporated in the crude oil export price return series such as volatility cluster,
autoregressive conditional variance and in modeling Heteroskedasticity (ARCH) effect of
both symmetric and asymmetric terms. This also take into consideration high persistence of
volatility which is one of the crux for models that have the most powerful ability for
charactering change in volatility. According to Olugbenga et al (2017), the ARCH model is
basically rooted in monthly return (usually squared returns) of stock prices.
Olugbenga et al, (2017) further opined that an ARCH model is a stochastic process with
autoregressive conditional heteroskedasticity. It is a simple model that can capture or
described a stochastic process which could either be non-stationary but asymptotically
stationary. If the stochastic process shows clustering volatility, then the ARCH models can
be applied.
Although, due to the shortcoming as a result of weak signals about the level of volatility led
Belleslev to develop an extended version of the ARCH model refers to as the Generalized
Autoregressive Conditional Heteroskedasticity (GARCH) (Olugbenga et al, 2017). The most
interesting aspect of this model (GARCH) is that even with a small number of terms it
appears to perform better than an ARCH model with many terms. The general belief about
the model is that during a period of decline in growth, volatility is likely to increase and
during a period of increasing growth, it is likely to decline as used in most studies.
Autoregressive conditional heteroscadasticity and its variances extensions are Generalized
ARCH, GJR-GARCH (TGARCH), Exponential GARCH, component – GARCH, and
GARCH-M. In some cases, the first-order GARCH family models have been extensively
proven to be appropriate for modeling and forecasting financial time series as observed by
Hsieh (1991), Bera et al (1993), Hansen et al (2004), Eric (2008), Olowe (2011), Hojatallah
et al (2011). But in these studies little or no attention has been given to their suitable error
distribution assumptions for modeling especially the normality, student’s-t and generalized
error distribution assumption.
This is considered because review of relevant literature shows that several researchers have
neglected the contribution of the error distribution assumptions while modeling market price
volatility. The wrong use of an appropriate error distribution in volatility model for financial
time series may cause misspecification in volatility model, leptokurtic and autocorrelation
IIARD – International Institute of Academic Research and Development
Page 28
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
behavior of such series. Whereas Klar et al., (2012) posited that in appropriate specification
of the concept distribution may lead to a sizeable loss of correctness of the corresponding
estimators, wrong risk determination, inaccurately priced options and inadequate assessment
of value-at-risk (VAR). In modeling volatility these is need to specify the form of the error
distribution to be used in the estimation. Hence, this study seeks to investigate and as well as
close gap the vacuum in several literatures by using the three commonly used first order
symmetric GARCH family models on the form students-t, normal (Gaussian) and generalized
error distribution (GED) with a view to compare them to when it is used in asymmetric
GARCH family models, while considering the best fitted model for forecasting volatility with
the best error distribution for crude oil export price within the years under consideration.
3.0 Methodology
This Paper uses two steps estimation Procedure for Modeling Volatility.
a. The Time series/ Statistical Approach and
b. The Statistical / Econometrics Approach
3.1. The Time series/ Statistical Approach: This involves the time plot and Normality test.
According to Etuk (2017) time series is a set of data collected sequentially in time and
therefore such data have the characteristics of being correlated rather than being independent
like other data are assumed to be. Therefore, it is particularly important to draw a time plot to
examine the trend in their relationship graphically within the years under the study.
Normality Test is conducted on the transformation of return of crude oil price using the
Jarque-Bera test statistics. According to (Chinyere, Dickko and Isah, 2015) Jargue-Bera is
defined as joints test of skewness and kurtosis that examine whether data series exhibit
normal distribution or not; and this test statistic was developed by Jargue and Bera (1980). It
is defined as;
X
2
~
2
N 2 K 3
S
6
4
(3.23)
Where S represents Skewness, K represents Kurtosis; N represents the size of the
macroeconomic variables used. The test statistic under the Null hypothesis of a normal
distribution has a degree of freedom 2.When a distribution does not obey the normality test
Abdulkarem et al (2017) suggests that the alternative inferential statistic for such especially
the case of error distribution assumptions with fixed degree of freedom are fussed into the
ARCH and GARCH models.
3.2 Statistical/ Econometric Approach: This shall be done under the following procedure:
Testing for ARCH (1) affects, Model Estimation with Symmetric and Asymmetric GARCH
Model, Model Diagnostic.
3.2.1 Testing for ARCH (1) effects
This test is carry out to examine the presence of heteroscedascity in the residuals of crude oil
export prices, and this is done using LaGrange multiplier (ML) as proposed by Engle (1982).
This is done by obtaining the residuals first from the ordinary least squares regression of the
conditional mean equation which might be represented in (ARMA) process. Supposing an
ARM (1,1) process is considered for example, if the conditional mean equation is
Xt = 1Xt + t + 1t-1
IIARD – International Institute of Academic Research and Development
(3.24)
Page 29
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
Once the residuals t is obtained, then the squared residual on a constant and its q lags as in
the following equation:
t2 0 i t21 .... q t2 q 0 i t21
q
11
t 0 i t21 .... q t q t 0 i t 1 t
(3.25)
The ARCH model (q) is
q
11
Where;
(3.26)
t is the unconditional variance
0 is the constant term
i is the co-efficient of the ARCH term
t-i is the corresponding lags of the errors at time t-1
q is the length of ARCH lags and
t is the error term
The Test Hypothesis:
Ho: 1 = …. = q = 0 (Absence of ARCH effect up to order q) Against H1: 0 for some i ∑
{1,….q} at least one variable has presence of ARCH effect. The number of observations
times the R-squared (nR2) gives the test statistics for the joint significance of the q – lagged
squared residuals with q degrees of freedom. Hence, nR2 is tested against X2(q) distribution.
If nR2> X2(q) table result, then the Null Hypothesis will be rejected and conclude that there is
an ARCH effect in ARMA model.
3.2.2 Model Estimation Using Symmetric and Asymmetric GARCH Models:
In line with the objective of the study, the model adopted for the study was derived as thus:
Supposing we have a regression model given as;
Yt = K1X1t + t
(3.1)
Where t is the residuals, then
(3.2)
t
= ht Z t
ht 0 i t21 B j t j
9
i 1
j 1
(3.3)
While o is a constant term, i is the co-efficient of 1 is the elasticity coefficient and t is
the stochastic disturbance term. It is important to note that, for equation (3.1) and (3.3) to
exist, them; o, i, i-t,j > 0
However, GARCH (p,q) is model as thus: The mean is written as
Xt = + xt-i + t, t N(0,t)
(3.4)
t2 0 i t 1 y j t j
Where; the variance component is written as thus
q
i 1
j 1
(3.5)
The mean equation (3.4) become a standard, for other models alongside with conditional
variance components, when p=1 and q=1 then it is consider as a case of GARCH (1,1).
Where all the parameters o, i, yj, 0; t2 is the conditional variance, o, j constant term,
IIARD – International Institute of Academic Research and Development
Page 30
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
j and yj are coefficients of the ARCH and GARCH term respectively, t2i and t2 j are the
squared errors at lagt-i and t-j respectively.
Equation (3.1), (3.2), (3.3) (3.4) and (3.5) provide a priori expectation expected signs and the
significant of the value of the co-efficient of the model parameters to be estimated in light of
economic theories and empirical evidence. Equation (3.5) is defined as GARCH (p,q) model
an extended framework of ARCH(q) model as proposed by Bolleslev (1986) in which it is
refers to as the P lags of past conditional variance. The GARCH (p,q) with Zt as a discrete
time stochastic process is defined as:t = Ztt and is weakly stationary with
E(t) = 0 and
q
Var (t) = 0 1 X i y j
j 1
l 1
Cov (t, t) = 0 for t s if and only if
1
y
i 1
q
i
j 1
j
(3.6)
1, (αo> 0), for the system be stationary
Also, GARCH in MEAN (GARCH – M) Model as propose by Engle et al (1987) mostly
estimate return of financial data series as dependent of the conditional variance of a standard
deviation. It model high risk that often accompany high expected return. The simplest form of
GARCH-M model is the GARCH-M model is the GARCH-M (1, 1) written as:
Mean equation: Xt = + 2t + t, t~ N (0,t2)
(3.7)
Variance equation;
2 = α0 + α1, 2t-1 + β12t-1
(3.8)
Similarly, the Threshold Generalized Autoregressive Conditional Heteroskedasticity Model is
generally specified in its conditional variance using the acronym TGARCH (p,q) and it is
written as thus;-
0 i
q
2
t
i 1
2
t 1
yi I t i
q
i 1
2
t 1
j te j
j 1
(3.9)
Where It-i = 1 if t2 < 0 and zero. In equation (3.7), Goods news implies that t2 > 0 whereas
bad news implies that t2 < 0 under these conditions, (shocks) of equal magnitude have
differential effects on the conditional variance. However, good news has an impact of i
white bad news has an impact of i + yi. Bad news increases volatility if yi> 0 which
invariably means that there is existence of leverage effect in the ith order when yi 0 them the
news impact is asymmetric. But, the first order representation of the equation is given as
TGARCH (p,q) is;
t2 0 1 i t21 y1 I t i t21 j te j
q
i 1
(3.10)
The Exponential Generalized Autoregressive Conditional Heteroskedasticity (EGARCH)
model as proposed by Nelson (1991). The conditional variance of EGARCH (p,q) model is
specified generally as;
IIARD – International Institute of Academic Research and Development
Page 31
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
Log
t i
B j Log t2 j
(3.11)
0
t 1 j 1
t-i> 0 and t-j< 0 depicts good news and bad news respectively, whereas their total effects are
t i
yi
i
t 1
i 1
q
2
t
given as 1 yi t i , and 1 yi t i . When yi< 0, the expectation is that bad news
enhances volatility persistence to be high. The EGARCH model achieves covariance
stationarity when B j 1
j 1
However, the target of this study is to model the conditional variance using EGARCH (1,1)
model which would be refine as
1
Log t2 0 1
(3.12)
y1 t 1 log t21
t 1
t 1
The total effects of good and bad news for EGARCH (1,1) are 1 yi t i and 1 yi t i
respectively. We accept the Null Hypothesis that yi = 0 shows the presence of leverage effect,
i.e. bad news have stronger effect than good news on the volatility of the return series.
And the power Generalized Autoregressive Conditional Heteroskedasticity(P GARCH)
model as proposed by Ding et al (1993) expressed conditional variance using P GARCH
(p,d,q) as;
td 0 i t i Yi t i B j td j
q
11
d
q
J 1
(3.13)
The failure to accept the null Hypothesis Y1, 0 shows the presence of leverage effect. In
order to ensure that the all the model used here efficiently capture the characteristic of
Gaussian and non-Gaussian process for high volatility in financial time series equation (3.5),
(3.7), (3.8, (3.9),(3.10) and (3.11) above were subjected to error distributional assumption as
specified .
Firstly, the normal distributional assumption; this assumed that the variance in the entire
GARCH model given above utilizes the likelihood function of their residuals and variance.
1 T
2
(3.14)
Lt In2 In t2 t2
2 t 1
t
t2 is Specify field in each of the GARCH models.
Similarly, following the assumption that GARCH models follow generalized Error
distribution tends to account for the Kurtosis in series returns, which are not properly
captured using the normality assumption as shown in equation (3.12) above, the volatility
models are estimated with generalizes error distribution by maximizing the Likelihood
function below;
IIARD – International Institute of Academic Research and Development
Page 32
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
1
3 yt X 1
t
1
1
v
2
log t v
L t log
2
2
t2 1v
3 v 2
v v
v
2
2
(3.15)
Where V represents the shape of the parameter use in the estimation and this shows the
Skewness of the return series used in the estimation and V > 0. The higher the value of V, the
greater the corresponding associated with the weight of the tail. Generalized Error
distribution (GED) reverts to normal distribution if V = 0. And finally in the case of the
student’s distribution the volatility models here are estimated to maximize the likelihood
function of a student’s t distribution;
2
2
r
H Yt X t1
1
( ) 2 1
2
(3.16)
L t
log
2 log t 2 log1 2 2
2
2
t
H
2
Where is the degree of freedom that controls the behaviour of the tail.> 2.
3.3 Nature and Sources of Data
Data used for this study was sourced for from the central Bank of Nigeria (CBN) statistical
database website (www.cbn.gov. ng). The variables comprised of monthly crude oil export
prices (COP), extracted from the month of January, 1987 – June, 2017. These make a total of
366 data points. Crude oil export prices conditional variance models are fitted to
conditionally compound monthly return computed as,
COPRt =
log
COPt
*100
COPt 1
(3.17)
For t = 1, 2, ….t-j where COPRt is the crude oil export price return at time t, COPt is crude
oil export price at time t and COPt-1 is crude oil export price at time “t-1’’ .
The variable was well differenced (D) to get rid of outlier and as well obtain stationarity
within them. The data was analysis using Eviews Software version 9.
IIARD – International Institute of Academic Research and Development
Page 33
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
Results and Discussions
4.1 Time plot / Test for Volatility clustering
140
Crude Oil Price (US&Barrel)
120
100
80
60
40
20
0
88
90
92
94
96
98
00
02
04
06
08
10
12
14
16
Year
Figure 4.1: Monthly Price of Nigeria Crude Oil Export Market (US Dollar/Barrel) –
From January, 1987 to June, 2017.
Figure (4.1) illustrates the dynamics of crude oil prices series. The behavior of crude oil
prices from January, 1987 to June, 2017 and this reveal an upward trend which later falls
within the year 2014-2016.
Test for Volatility clustering
50
Crude oil Price Return(COPrt)
40
30
20
10
0
-10
-20
-30
-40
88
90
92
94
96
98
00
02
04
06
08
10
12
14
16
Year
Figure 4.2: Monthly Price Return of Nigeria Crude Oil Export (US Dollar/Barrel) –
From January, 1987 - June, 2017.
IIARD – International Institute of Academic Research and Development
Page 34
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
Figure 4.2 above, clearly show evidence of volatility clustering in the returns series of crude
oil export price US dollar/Barrel and the crude oil export price exhibit sharp increase with a
corresponding sharp decrease.
4.2 Descriptive Statistics of Crude oil price Return Series.
This is done to tested normality and to examine whether the variable under the study is useful
for analysis
Table 4.1: Summary Statistic of Crude oil Export Price Return
Mean
Median
Min
Maxi
Std.Dev.
Skewness
Kurtosis
JarqueBera
Prob. Value
0.002594
0.005365
0.321046
0.470843
0.089797
-0.131519
5.343272
84.56002
0.000000
Source: Researcher’s Computation, 2017. It is all tested Significant at 1 and 5%
respectively
Table 4.1 shows the descriptive statistic for the data variable and its return series covering the
period of January, 1987 –June, 2017. The Margins between the minimum and maximum
values of the series indicate evidence of variability of the trend of the series within the period
under coverage.
4.3 Test for ARCH effect
Table 4.2: Estimation Results for Test for ARCH Effect
Heteroskedasticity Test: ARCH
Lag 1
F-statistic
13.39122
Prob. F(1, 361,5, 353,)
0.0003
n*R2
12.98377
X2 (1,5,)
0.0003
Source: Researcher’s Computation, 2017.
respectively
It is all tested Significant at 1 and 5%
Both the F-statistic and n*R2 test in table 4.2 indicate the existence of ARCH effect on an
increase in the variable even at 1% level of Significance for the first order autoregressive
process. The test for higher order lags is neglected reasoning been that Lag one test is
adequately enough for the modeling of volatility models considered in the study. The First
Order Symmetric GARCH Family Models In Error Distribution Assumption is estimated as
specified in equation (3.4), (3.5), (3.7), (3.8) and (3.9) in their specific error distribution
assumption as in equation (3.12), (3.13) & (3.14).
IIARD – International Institute of Academic Research and Development
Page 35
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
Table 4.3: Estimation Results of the First Order Symmetric GARCH Models in Error
Assumption Distribution.
Models
Equations Model Parameter
Intercept
Mean GCOP(-1)
Intercept
ARCH
GARCH (1, 1) Variance GARCH
AIC
Student’s
–tGeneralized Error Model
Error
Distribution
with
Distribution
Minimum
AIC
&SICAcrossE
rror Distr
Normal Error
Distribution
Coefficients
0.001443
0.174167
0.000701
0.259159
P-Value
0.7188
0.0031
0.0383
0.0000
Coefficients P-Value
0.003040 0.4384
0.168070 0.0030
0.000542 0.0883
0.191681 0.0031
Coefficients
0.002404
0.169706
0.000630
0.229175
P-Value
0.5518
0.0035
0.0777
0.0003
0.680844 0.0000 0.755009 0.0000 0.713596 0.0000
-2.107316
-2.121125
-2.108618
-2.121125
-2.056886
-2.053784
-2.056886
-2.044380
SIC
0.94669
0.942771
ARCH+GARCH 0.940003
SQRT(GARCH) 0.183998 0.4079 0.179470 0.3952 0.171953 0.4257
Mean
Intercept
GCOP(-1)
GARCH (1, 1)M
Variance Intercept
-0.012147 0.4708 -0.010320 0.5261 -0.010405 0.5299
0.176889 0.0028 0.168910 0.0029 0.172616 0.0030
0.000745 0.0328 0.000531 0.0894 0.000635 0.0736
ARCH
0.265656 0.0000 0.194859 0.0029 0.232898 0.0003
GARCH
0.668938 0.0000 0.754003 0.0000 0.709608 0.0000
AIC
-2.104650
-2.118279
-2.105658
-2.118279
SIC
-2.040411
-2.043334
-2.030713
-2.043334
0.948862
0.942506
ARCH+GARCH 0.934594
Source: Researcher’s computation, 2017. It is all tested Significant at 1 and 5%
respectively
The First Order ASymmetric GARCH Family Models In Error Distribution Assumption is
estimated as specified in equation(3.9),(3.10) and (3.11) in their specific error distribution
assumption as in equation (3.12), (3.13) & (3.14).
IIARD – International Institute of Academic Research and Development
Page 36
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
Table 4:4: Estimation Results for Asymmetric First Order GARCH Family Models in
Error Distributional Assumptions.
Model(s)
Equation(s Model Parameter(s)
Student’s –t
Distribution
Normal Error
Distribution
ErrorGeneralized
Distribution
ErrorModel
with
Minimum AIC &
SIC Across Error
Distr
Intercept
Coefficients
0.000667
P-Value
0.8751
Coefficients P-Value
0.001938 0.6276
Coefficients
0.001611
P-Value
0.6959
GCOP(-1)
0.178949
0.0032
0.179008 0.0016
0.175576
0.0030
Intercept
0.000650
0.0490
0.000496 0.1053
0.000580
0.0909
ARCH
0.210151
0.0002
0.123663 0.1139
0.177659
0.0150
Asymmetric
GARCH
0.067327
0.700169
0.4524
0.0000
0.091472 0.3337
0.778623 0.0000
0.068794
0.733782
0.4772
0.0000
AIC
-2.102914
-2.118247
-2.104302
SIC
-2.038675
-2.043301
-2.029357
Mean
TGARCH (1, 1)
Variance
-2.118247
-2.043301
0.911441
ARCH+GARCH 0.91032
Mean
EGARCH (1, 1)
Intercept
GCOP(-1)
Intercept (C(3))
ARCH(C(4))
Asymmetric(C(5))
Variance
GARCH (C(6))
0.902286
0.000996 0.8050
0.001330 0.7361
0.001499
0.7069
0.165820
-0.917644
0.437900
-0.068282
0.884734
0.170435
-0.727631
0.349107
-0.080388
0.908970
0.164921
-0.842727
0.400559
-0.070817
0.893938
0.0040
0.0088
0.0001
0.2056
0.0000
0.0038
0.0022
0.0000
0.1625
0.0000
0.0024
0.0162
0.0015
0.1671
0.0000
AIC
-2.118714
-2.128986
-2.117669
-2.128986
SIC
-2.054475
-2.054041
-2.042723
-2.054475
1.258077
ARCH+GARCH
1.294497
1.322634
Mean
PARCH (1,1, 1)
Intercept
0.000950 0.8113
0.001427
0.7191
0.001541
0.6990
GCOP(-1)
0.154498 0.0054
0.170937
0.0018
0.163322
0.0034
Intercept(C(3))
0.006899 0.0681
0.006033
0.1102
0.006677
0.0992
0.210965 0.0000
0.173097
0.0029
0.194785
0.0005
Asymmetric(C(5))
0.227742 0.1000
0.264919
0.2065
0.234415
0.1826
GARCH(C(6))
0.758050 0.0000
0.795798
0.0000
0.772165
0.0000
AIC
-2.113325
-2.124958
-2.113101
-2.113325
SIC
-2.049086
-2.050013
-2.038156
-2.050013
0.969015
0.968895
1.00658
Variance ARCH(C(4))
ARCH+GARCH
Source: Researcher’s Computations, 2017. It is all Tested Significant at 1 and 5%
respectively
IIARD – International Institute of Academic Research and Development
Page 37
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
Model Fitness & Selection
From the fifteen models (symmetric and Asymmetric) estimated above, models were selected
on the basis of Schwarz information criterion (SIC) as supported by Alhassan et al (2016) in
order to select the best model for forecasting. The results are presented in the table below:
Table 4.5: Model Fitness and Selection
First
Order Error Distributional Assumptions
GARCH Family Normal
Student’s-t
Generalized
Model.
Error Distr. Error Distr.
Error Distr.
GARCH(1,1)
-2.053784
-2.056886
-2.044380
GARCH-M(1,1)
TGARCH(1,1)
EGARCH(1,1)
PARCH(1,1,1)
-2.040411
-2.038675
-2.054475
-2.049086
-2.043334
-2.043301
-2.054041
-2.050013
-2.030713
-2.029357
-2.042723
-2.038156
Minimum
SIC
-2.056886
-2.040411
-2.043301
-2.054475
-2.050013
Source: Researcher’s Computation, 2017
Conclusively, the best fitted selected model are written as shown below: For the first order
Symmetric GARCH Model in student’s-t error distribution
Mean Equation:
COPRT = 0.00304 + 0.16807*COPRT (-1)
Variance Equation:
GARCH = 0.00054 + 0.19168*RESID (-1)^2 + 0.75501*GARCH(-1)
Similarly, for the first order Asymmetric GARCH Model we have EGARCH in normal as
given below:
Mean Equation:
COPRT = 0.00010 + 0.16582*COPRT (-1)
Variance Equation:
LOG(GARCH) = -0.91764 + 0.43790*ABS(RESID(-1)/@SQRT(GARCH(-1)))
0.06828*RESID(-1)/@SQRT(GARCH(-1)) + 0.88473LOG(GARCH(-1))
IIARD – International Institute of Academic Research and Development
-
Page 38
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
Parameter Estimation of the Selected GARCH Family Models
Table 4.6 presents the impact of news on volatility of crude oil export price in the best fitted
asymmetric volatility models, and their volatility persistence arising from the parameter
estimates of the five best models.
Table 4.6: News Impact Assessment and Test Volatility for Persistence
Parameter Estimates Asymmetric
GARCH
Family Symmetric
GARCH
of GARCH
Models
Models
TGARCH EGARCH PGARCH GARCH GARCH-M
Distributional
Student’s-t Normal
GED
Assumptions
Good News
0.123663
0.437900
Bad News
0.215135
0.369618
Volatility Persistence 0.902286
1.322634
Source: Researcher’s Computation, 2017
0.173097
0.438016
0.968895
0.94667
0.948862
Model Diagnostics
In order to ascertain the fact the selected models are good enough for forecasting, there is the
need for further confirmatory test and this shall test for the presence of ARCH effect, serial
correlation and Q-Q plots for the residuals using the selected models,
Test for ARCH Effect: This is done in conformity with the residuals of the models as
review in the concept of homoscedasticity as account for, in Arch effect model. This was
estimated using the ARCH –LM model and the results are shown below:
Table 4.7: Heteroskedasticity Test for the Five Best Fitted GARCH Family Model
Models
Heteroskedasticity
Lag 1
Lag 5
Lag 10
Test: ARCH
GARCH(1,1) in Student’s-t F-statistic
0.683883 0.498243 0.453985
Error Distribution
Prob. F(1,1234)
0.686371 2.515806 4.624236
GARCH-M(1,1) in Student’s-t F-statistic
0.652177 0.490809 0.476347
Error Distribution
Prob. F(1,1234)
0.654607 2.478527 4.848895
TGARCH (1,1) in student’s –t F-statistic
0.257402 0.403268 0.409138
Error Distribution
Prob. F(1,1234)
0.258644 2.038968 4.172819
EGARCH(1,1) in Generalized F-statistic
0.090917 0.341227 0.381143
Error Distribution
Prob. F(1,1234)
0.091398 1.726787 3.890427
PARCH(1,1) in Student’s-t F-statistic
0.720752 0.414262 0.413184
Prob. F(1,1234)
Error Distribution
0.723301 2.094228 4.213588
Source: Researcher’s Computations, 2017
5.1 Discussion of Results
The monthly crude oil price data for this study spans from January, 1987 – June, 2017 with
the total data points of 366, conditional variance models were fitted to continually, compound
monthly exchange rate. Fifteen models (15) were estimated using the first order GARCH
family model in its three error distribution assumptions. In the estimation of the models,
certain conditions were taken into considerations and this incorporate the pattern as shown by
the variable. These include the following: Time series plot, Descriptive statistic, Test for
ARCH effect test, GARCH family model Estimation and Model diagnosis test.
IIARD – International Institute of Academic Research and Development
Page 39
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
In the estimation as shown in Figure (4.1) illustrates the dynamics of crude oil prices and its
return series. The behavior of crude oil prices from January, 1987 to June, 2017 and this
reveal an upward trend which later falls within the year 2014-2016. Also, Figure 4.2 above,
clearly show evidence of volatility clustering in the returns series of crude oil export price US
dollar/Barrel and the crude oil export price exhibit sharp increase with a corresponding sharp
decrease. This also shows that crude oil export return price US dollar per Barrel has not been
actually stable within the sample period under this study. The return series follow an
unsteady pattern and the returns series confirmed that there is an evidence of volatility
clustering. This is also supported by Abdulkareem et al (2016) findings. The period of high
volatility, accompany with period of relative calmness the preliminary investigations show
that the variable exhibit unusual fluctuation using time series plot then after transformation
the trend in the graph became stationary with an increasing volatility clustering.
In another development, the variable was subjected to descriptive test for normality and the
result shows that the variable violates all the characteristics of variables that are normally
distributed. Table 4.1 shows the descriptive statistics for all the variables and their return
series covering from January, 1987 – June, 2017.The mean (0.002594) have positive signs,
meaning it is mean reverting. The standard deviation (0.089797) measure the riskiness of the
series under the study. The Higher the standard deviation, the increase in volatility of the
crude oil prices return and the risky the investment in this trade. The 8.09797% difference
between minimum and maximum return series is a clear evidence of the level of price
variability in fairness to trading in crude oil market within the sample period. Again, the coefficient of skewness -0.131519 is less than zero indicated that the distribution is negatively
skewed which one of the common characteristics of fairness in crude oil price return series
while the Kurtosis (5.343272) is greater than three (3). However, the Kurtosis of a normal
distribution is 3 which mean the distribution not normal. And the Jarque-Bera (84.56002)
accomplish with a very small corresponding probability value (0.000000), the Null
Hypothesis of Normality is rejected and the alternative inferential statistic as suggested by
Abdulkarem et al (2017) become necessary with their corresponding error distribution
assumptions and fixed degree of freedom fussed into the ARCH and GARCH models .
A look at the table (4.3) reveals the values of F-statistics (13.39122) to be higher with its
corresponding chi-squares statistics less than the Obs. R-squared (nR2) (12.98377) i.e. the
Obs. R-squared is greater than prob. Chi-square. Hence, the Null hypothesis is rejected
therefore it can be concluded that there exist ARCH effect in crude oil export price return
series, even when it was tested at 1% significance level. See complete estimation results for
the test for ARCH effect in appendix. This confirmed Abdulkarem et al (2017) assertion
about variables that can be estimated using GARCH family model
Table 4.4 and table 4.5 presents comprehensive analysis on crude oil export price in dollars
per Barrel while selection were done only with the model with the least Schwartz information
criterion. The symmetric models in the table4.4 above reveal that all the ARCH Coefficients
in the three error distribution assumption are statistically significant at the 5%levelof
significance. This evidently confirmed the presence of ARCH effects and this support the fact
that the previous month’s crude oil export price information can actually influence the
present month crude oil export price return. That is crude oil export price volatility is
influence by its own ARCH and GARCH.
Similarly, it is clear that @ SQRT (GARCH) coefficients are not significant and it does not
provide much needed information on the volatility of return series. However, the results in
IIARD – International Institute of Academic Research and Development
Page 40
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
GARCH (1,1) and GARCH-M (1,1) shows that the sum of the ARCH and GARCH
coefficients are less than one. This indicates that using GARCH (1,1) and GARCH-M(1,1) in
modeling characteristics exhibited by volatility of crude oil export price within the sample
period reveal a mean reverting condition.
Also, considering the degree of effect or persistence in GARCH (1,1) according to the order
their of error distribution assumptions such Normality, student’s-t and the generalized error
assumptions. The GARCH (1,1) in Normal error distribution have (94.003%), GARCH(1,1)
in Student’s-t gives have (94.669%) and GARCH(1,1) in Generalized error distribution have
(94.2771%). This follows that GARCH (1,1) in Normal error distribution have the highest
volatility persistence, follow by GARCH (1,1) in student’s-t and GARCH (1,1)in generalized
error distribution. Meanwhile ,the degree of effect or persistence in GARCH-M(1,1) are as
follows: GARCH-M(1,1)in Normal error distribution is (93.4594%), GARCH –M (1,1) in
student’s-t (94.8862%) and GARCH-M(1,1) in Generalized error distribution is (94.2506%).
This shows that using GARCH-M(1,1) in modeling volatility, GARCH-M(1,1) in normal
error distribution have the highest level of volatility persistence or effect, follow by the
GARCH-M(1,1) in Generalized error distribution and the GARCH-M (1,1) in student’s-t
distribution. Using the GARCH-M (1,1), it shows that increased risk leads to a higher return.
Finally, comparing the two models on the basis of fitness and performance using the
Schwartz information criteria, GARCH(1,1) in student’s error distribution assumption has the
value (-2.056886) with the Akaike information criteria(AIC) of -2.121125 and GARCHM(1,1) in student’s-t error distribution (-2.043334) with the Akaike information criteria(AIC)
(-2.121125 ) were chosen as the best fitted symmetric models for estimating crude oil export
prices within the sample period.
Based on the results of the findings the symmetric GARCH models in student’s-t error
distribution clearly perform than the asymmetric GARCH models. This is also confirmed in
Shamiri and Isa (2009) findings while modeling and forecasting volatility of the Malaysean
stock markets. According to the Schwartz information criterion, GARCH (1,1) in student’s-t
error distributional Assumption in symmetric GARCH and EGARCH (1,1) in normal error
distributional assumptions in Asymmetric GARCH outperform other models irrespective of
their class. Although, one unique behavior about GARCH-M (1,1) model is that it allow
conditional mean of a financial data return sequence to depend on its conditional standard
deviation or variance.
Similarly, in evaluating GARCH family models performance in Nigerian crude oil markets
there is the need to do a comparative volatility modeling of Nigeria crude oil using symmetric
and Asymmetric GARCH models. This will help us in drawing conclusion about the best
fitted model. The asymmetric first order GARCH family models in error distributional
assumptions in the equation (3.10),(3.11) and (3.12) were also estimated using the residual
from equation (3.23) for each of the model in their error distribution assumptions. Also, the
value of the Power GARCH otherwise refers to as the PGARCH (1,1,1) inputted according
their order of degrees. This evaluated alongside with other two models generated nine
volatility models as shown in the table 4.4 above.
In the table 4.4 above, the entire ARCH (α) coefficient in all the models shows positive sign
and they are statistically significant at the 5% level of significance except the case of
TGARCH in student’s–t whose probability value (p-value) is 0.1139. These confirmed the
presence of ARCH effect. It also imply that the previous month’s crude oil export price return
IIARD – International Institute of Academic Research and Development
Page 41
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
series information influence this presence month’s crude oil export price return series.
However, in spite of such deduction one can equally say that there exist a leverage effect
and this means bad news can have impact on conditional volatility than good news. But for
TGARCH in student’s–t error distribution models, the above implications is a contradiction
simply because the non-significance of the p-value, mean that the previous month’s volatility
can’t influenced this present month’s volatility. Also the asymmetric term has positive coefficient but not statistically significant at the 5% level of significance. This is also a
confirmation to (Abdulhakeen et al, 2016) findings.
But the asymmetric terms in the other models have negative signs in their co-efficient but
they are not statistically significant at 5% level of significance. This means that there exist
negative correlation between the past crude oil export price return series and future volatility
of the return series. And that the higher leverage effect that occurs due to negative crude oil
export prices return will likely translate to low equity price otherwise result in sky rocking of
debt to equity ratio. Also, it can be deduced that negative shocks reduces the volatility of
crude oil export price returns and that negative return of crude oil has more impact on the
volatility of crude oil export price s than the positive return series.
Also, the addition of the ARCH and GARCH coefficients are very closed to unity i.e.
TGARCH(1,1) in normal error distribution (0.91032), TGARCH in student’s-t error
distribution (0.902286) and TGARCH(1,1) in Generalized error distribution (0.911441), then
this invariably means they are all mean reverting in nature.
And the degrees of shocks are permanent at 91.032%, 90.2286% and 91.1441% of
persistence respectively. This implies that using TGARCH(1,1) in modeling volatility within
the present sample period TGARCH(1,1) model in Generalized error distributions have the
highest volatility persistence(91.1441%) follow by TGARCH(1,1) in normal error
distribution( 91.032%) and the TGARCH in student’s-t error distribution(90.2286%) .
Whereas comparing the models on the basis of the Schwarz information criterion (SIC) and
Akaika information criteria (AIC),TGARCH(1,1) model in student’s-t error distribution have
the least value (-2.043301) with (AIC) (-2.118247) this was considered the best fit.
Similarly, The EGARCH (1,1) model estimation was also considered in the modeling and all
the co-efficient of the ARCH(α) terms show negative signs and significant at the 5% level of
significance. This simply means that there exists the presence of ARCH and leverage effect.
It is clear that the result reveal negative correlation between the past return of crude oil price
and future volatility. It also shows that bad news has more impact on the volatility of the
returns series (crude oil export price) than the positive news. Also, all the asymmetric terms
such as (-0.068282 in normal error distribution, -0.080388 in student’s –t, -0.070817 in
Generalized) has negative signs but they are not statistically significant at 5% level of
significance. This reveals that negative shocks reduce the volatility of the variable than the
positive shocks of the same magnitude whereas leverage effects are considered necessary on
the basis of EGARCH (see Abdulhakeem et al, 2016).
The sum of the ARCH and GARCH has the following values; EGARCH (1,1) in normal error
distribution assumption (1.322634), EGARCH(1,1) in student’s–t error distribution
assumption(1.258077) and EGARCH in Generalized error assumption distribution is
(1.294497). These simply mean that all the EGARCH models are mean reverting and their
volatility persistence are only temporary. Although, the degree of their persistence are high,
ranging from 132.2634% to 125.8077% and 129.4497% respectively.
IIARD – International Institute of Academic Research and Development
Page 42
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
Similarly, the Schwarz information criterion and Akaike information criteria is given that
EGARCH(1,1) in normal error distribution (-2.054497) while (AIC) -2.128986 in
EGARCH(1,1) in student’s-t .However, EGARCH(1,1) in Normal Error distribution was
considered the best fitted since the model have the least Schwartz information criterion and
according to Alhassan et al(2016) SIC is use as the best choose for selection of model fitness
since it levies heavy penalties for loss of degree of freedom.
Finally, the PGARCH(1,1) was also considered for the study and the results from the
estimation as shown in table (4.4)above indicate that all the ARCH(α) co-efficient in the
models has positive signs (0.210965, 0.173097,and 0.194785) and they are statistically
significant at the 5% level of significance. This reveals that the models have ARCH and
leverage effects. Hence, we can say that there exist negative correlation between the first
return of the series and future volatility. Also, this shows that negative news have more
impact on the volatility of return than the positive news. Furthermore, all the asymmetric
terms have positive signs (0.227742, 0.264919, and 0.234415) but they are not statistically
significant at the 5% level of significance and this confirmed the fact that there exists
leverage effect. Although, the leverage effect is not necessary considered on the basis of
PARCH modeling. Also, the sum of the ARCH and GARCH terms estimated at 0.969015,
0.968895 and 1.00658. These reveal that the models are mean reverting with persistence
shocks. The degrees of volatility persistence are in the following order PGARCH in
Generalized (100.658%), normal (96.9015%) and student’s-t error distribution (96.8895%).
In the results, the TGARCH(1,1),EGARCH (1,1)and PARCH(1,1,1) indicated the existence
of leverage effect in the market. This invariably means that bad news have much effect on
subsequent period volatility than good news of the same magnitude. The GARCH(1,1) in
Student’s–t error distributional assumption for symmetric first order GARCH family model
and EGARCH (1,1) in normal error distributional assumption were considered the
appropriate model since they were able to meet the model selection criterion. GARCH(1,1)
having the ARCH and GARCH summations mostly less than one whereas the later having
greater than one i.e. making ensure that stationarity of the model, been able to capture
leverage effect and its ability in eliminating ARCH effects.
Model fitness and selection are done as reveal in table 4.5. In table 4.5, GARCH and
GARCH-M in student’s-t error distribution were considered best fitted symmetric models
since they have the least Schwarz information criterion across the models while in the
asymmetric GARCH models, TGARCH in student’s–t, EGARCH in normal and PGARCH in
student’s-t error distribution assumption were considered the best fitted. However, GARCH
(1,1) in student’s-t error distribution assumption with SIC (-2.056886) was considered the
overall best fitted. In view of the above, the overall best fitted model is GARCH (1,1) model
in student’s-t error distribution for symmetric follow by EGARCH in normal distribution for
Asymmetric GARCH family model.
Our overall selected EGARCH model in normal error distributional assumption result
suggested that there exist a comparatively petite volatility in the crude oil market, although
on a diminutive level with a price fluctuation approximately at the level of 0.002594
USD/barrel for Nigerian crude oil which is also in conformity with Morardand Balu (2014)
findings.
Table 4.6 talks about new impact assessment and test for volatility persistence. Although,
these are carried out using the parameters estimated with respect to the selected GARCH
family model. The results of the parameter estimate of the selected first order asymmetric
IIARD – International Institute of Academic Research and Development
Page 43
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
GARCH family model in table 4.6 vividly show that bad news actually have more impact on
volatility than good news. Also, it is revealing that EGARCH in normal error distribution has
the highest overall volatility persistence. These models were diagnosed to ascertain whether
the ARCH effect (Heteroscedacity) have been actually eliminated using serial correlation test
and QQ–plot of the residuals of the models. From the results in table 4.7 the hypothesis that
there is no Arch effect (Null Hypothesis) is confirmed at the 5% significant level.
This was carried out to confirm whether the models have serial correlation which is not good
enough for models that can be used in making forecast.
Conclusion
The results from the study have been amazing as marketers, and investors alike have clearer
views on how to go about their transaction. Also, the leverage crude oil market shown by
EGARCH model is statistically significant at 1% level with a negative sign, which reveal
that negative shocks meaning a higher next period conditional variance than positive shocks
with the same sign, showing that the existence of leverage effect is observed in returns of the
crude oil market index. Contrarily to the EGARCH model, the leverage crude oil market
shown by TGARCH is statistically significant with a positive signs, this reveal that at some
point in the market positive shocks meaning a higher next period conditional variance than
negative shocks with the same sign, showing that the existence of leverage effect is observed
in returns of the crude oil market index. In conclusion the economic slowdown in a promising
market like this is challenging. However, the new convention in the market represents new
challenges for economist, econometricians and researchers a like to urgently do structural
reform in adjusting to news in the context of modeling price Volatility in any markets.
Recommendations
In the words of Jin (2008), opined that volatility increases the risk and uncertainty of external
transactions and predisposes a country to volatility related risks.
Considering the level of risk that accompany external trade and investment in stocks and
price of commodities with its corresponding return series, investors, financial analyst and
Government are advice to be careful and such the following recommendations were
suggested as thus :
When modeling price volatility different error distributional assumptions should be
specifically incorporated into the system as incorrect error specification may lead to
incorrect estimation, which could cause loss of efficiency in the model.
Also, investors should not close their eyes to the impact of news while forming
prospect on investment as the higher the standard deviation in the descriptive statistic
of the return series maybe vulnerable risks.
Government should look for new ways to diversify the economy from total
dependence on oil and non-crude oil such as agriculture to explore other sectors like
the manufacturing sector to reduce price volatility in the economy and its overall
effect on other macroeconomic indicators.
Exchange rate between Nigeria and her foreign trading partners should be regulated
to currency variability which may in turn affect other Macroeconomic indicator
References
Abogam. O,P. Akimola K.B and Baruwa O.I (2014) Non-Crude Oil Export and Economic
Growth in Nigeria (1990-2011), International Journal of Research in Economics and
International Finances 3(1) PP1-11.
Abwaku E. Newman C.O, Caniyu K.S, Maaji, U.Y, Oladuni A, and Zainab S. (2011): Global
Oil Prices Shocks and the Nigerian External Sector: An Empirical Investigation.
IIARD – International Institute of Academic Research and Development
Page 44
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
Adenugba A. A and D.PO S.O (2013). Non-Oil Exports in the Economic Growth of Nigeria;
A Case Study of Agricultural and Mmaral Resources. Journal of Educational and
Social Research 3(2) 403-418.
Agenor P.R, McDermolt C.J and Prasad E. (2000). “Macroeconomic Fluctuations in
Developing Countries: some Stylized Facts” World Bank Economic Review 14,
2511286.
Akpan E.O (2009) Oil Price Shocks and Nigeria Macro Economy. A papers presented at the
Annual Conference of SCSAE Conference, Economic Development in Africa,
Oxford.
Anderson T.G and Bollerstev T. (1998).An severing, the skeptics: Yes standard variance
models do provide accurate forecasts. Int’l Econ Rev. 39; 885-905.
Angabini A. WasiuZaman S. (2011). “GARCH Models and the Financial Crisis.A Study of
the Malaysian Stock Market”. The International Journal of Applied Economics and
Finance, Vol. No. 3 Pp. 226-236.
Anthony E.A and Adejumo V.A (2014). Exchange Rate Volatility and Non-Oil Exports in
Nigeria; 1986-2008.
Anyanwu A. (2000). Research Methodology in Business and Social Services.(Anun
Publishers Nig. Ltd. Owerri.
Aristotelous K. (2001) “Exchange Rate Volatility, Exchange Rate Regime, and Trade,
Volume: Evidence from the UK-US Export Function (1989-1999) “Economics
Letters 72, 187-94.
Agbede Moses O. (2013). “The Growth implications of oil Price shocks in Nigeria. Journal
of emergently Trends in Economics and Management; Pg 343-349.
Abduikareem A. and Abdulhakeem K.A (2016).Analyzing Oil Price – Macroeconomic
Volatility in Nigeria.CBN Journal of Applied Statistics Vol. No 1 (a) (June, 2016).
Atoi M. (2014). Testing volatility in Nigeria Stock Marketing using GRACH model.CBN
Journal of Applied Statistics Vol. 5. No. 2 December, 2014 Pp. 65.
Azeez, B.A, Kalapo K.T and Ajayi L,B (2012) Effect of Exchange Rate Volatility on MacroEconomic.
Balaban E. (2002). “Comparative Forecasting Performance of Symmetric and Asymmetric
Conditional Volatility Models of an Exchange Rate” Working Paper University of
Edinbusgh Center for Financial Markets Research, Edinburgh Pp. 1 – 14.
Benabsy – Onere A, Mignon V, Period A (2007), China and the Relationship between the Oil
Prices and the Dollar. Energy Policy 35: 5795-5805.
Bera A.K. and Hagging M.L (1993).ARCH Models Properties, Estimation and Testing,
Journal of Economics Survey (4) 367-366.
Bernard A.B and Wagner J. (1997).Exports and Success in German Manipulating
WelwirtschaftlichesArchiv 133 (1), 134 – 157.
Bernard A.B and Jenson B. (1999). Exceptional Exporter Performance: Cause, effect of both?
Journal of International Economics 47 (1997) 1- 25.
Brooks, C. (2006). Introductory Econometrics for Finance. Second Edition, Cambridge
University Press.
Bernard O. (2014) “ Falling Crude Oil Prices and Nigeria’s Response”. The Gucirdum
Newspaper December 29, 2014.
Black J. (2002). Oxford Dictionary of Economics Second Edition, Published in the United
States by Oxford University Press Inc. New York.
Bollersehev .T (1986). Generalized Auto Regressive Conditional Heteroskedasticiy. Journal
of Econometrics 31, 307-321.
Brada J.C and Mendez J.A (1988). Exchange Rate Risks, Exchange Rate Regimes and the
Volume of International Trade KYKLOS 41, 263-280.
IIARD – International Institute of Academic Research and Development
Page 45
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
Bredin .D, Formtas S. and Murphy E. (2003).An Empirical Analysis of Short-run and Long
run Irish Export functions; does Exchange Rate Volatility matters? International
Review of Applied Economics 17, 193-208.
Chinyere S.E. Dikko. H.G and Isah, A (2015). Modeling the Impact of Crude Oil Price
Shocks on Some macro-economic Variables in Nigeria Using GARCH and VAR.
American Journal of Theoretical and Applied Statistics, Published Online August 17,
2015, http: www.Science.publishinggroup.com/j/ajtas).
Courage .M. and Kin .S. (2014), The Impact of Oil Prices on the Exchange Rate in South
Africa. Journal of Economics, Vol5(2) : 193-199 (2014).
Cushman D.O (1889). Us Bilateral Trade Flows and Exchange Rate Risk During the Floating
Period. Journal of International Economics 25, 317-330.
Cushman D.O (1983). The Effects of Real Exchange Rate Risk on International Trade.
Journal of International Economics 15,45-63.
Central Bank of Nigeria (2017), http://www.Statistics.cbn.gov.ng/statsonline (retrieved Nov.
16th April, 2017).
Cushman D.O (1986) Has Exchange Rate Risk Depressed International Trade? The Impact of
Third-Country Exchange Rate Risk. Journal of International Money and Financial
5.361-379.
Doyle E. (2001). Exchange Rate Volatility and Irish-UK Trade, 1979-1992. Applied
Economics 33,249-265.
Ederington L .H, Fernano C.S, Lee, and Anthony D , (2011) Factors Influencing Oil Prices:
A Survey of the Current State of Knowledge in the Context of the 2007-08 Oil Price
Volatility.
Econometrics Working Paper Ewpooo2.
Englama A. Duke, O. Ogunleye S. and Ismail F. U. (2010). Oil Price and Exchange Rate
Volatility in Nigeria: An Empirical Observation. Central Bank of Nigeria Economic
and Financial Review 48 (3): 31-48.
Engle R.F (1982). “Autoregressive Conditional Heteroskidaticity with Estimates of the
Various of UK Inflation” Economical Vol. 50 No. 4 PP. 987 1008.
Eric Z. (2008). Practical Issues in the Univariate GARCH models.Handbook of Financial
Statistics, Springer – Verlay.
Etherr W. (1973).International Trade and the Forward Exchange Market. American Review
63(3) 494-503.
Fischer S. (1981). Relative stocks, relative Price variability and inflation. Brookings Paper on
Economic activity 2: 111-137.
Farrell .G.N, Kahn B. and Visser F.J (2006).Price Determination International Oil Markets:
Developments and Prospects “SARB Quarterly Bulletin No.219 March 2006, 69-88.
Felix J.M and Anaele A. (2006).Research Methods in Social Sciences.NSSI.Publishing
Company Port Harcourt. ISSN: 978-46823-9
Folorunso S.A and Olajide A.J. (2016). Price Instability, Exchange Exchange Rate volatility
and the Nigerian Economy: An Empirical Analysis. Journal of Economics and
sustainable Development Vol. 7.No. 4 2016 ISSN 2222 – 2855 (online).
Giles J.A and Williams C.L (2000). Export-hed Growth: A survey of the empirical Literature
and non Causality Results part 2.
Gerardo E. and Felipe L.B (2002), “The Impact of G-3 Exchange Rate Volatility on
Developing Countries”.AG-24 Discussion Paper Series (16) On United Nations
Conference on Trade and Development (UNCTAD), Harward University.
Gujerati D.N (2005). Basil Econometrics (4th Edition) McGraw-Hill Publishing Company
Ltd, New Delhi 110008.
Hall R. and Hitch C. (1939) “Price Theory and Business “Export Economic Paper Volume 2
IIARD – International Institute of Academic Research and Development
Page 46
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
Pp. 12-45.
Hamuton J. (1983). Oil Macro-economy since World War II.Journal of Political.
Harsan V. (2012).Exchange Rate Volatility in Turkey and its Effect on Trade Flows, Journal
of Economic and Social Research 4(1).
Hooper r. &Kohlhagen S. (1978).The effects of Exchange Rate Uncertainty on Prices ad
Volume of International Economics 8,483-571.
Hassen P. and Linda A. (2004). Forecast Comparison of Volatility Models. Does anything
Beat a GARCH (1,1) model? Journal of Applied Econometrics; 20: 873-889.
Hojatallah G. and Ramanorrayarnan P. (2011): Modeling Asymmetric Volatility in the Indian
Stock Market. International Journal of Business and Management 6(3).
Hsieh D. (1999). “Chaos and Non-linear Dynamics: Application to Financial Market’s
Journal of Finance 46 1839-1877.
Imoughele L.E and Ismaila .M. (2015).The Impact of Exchange Rate on Nigeria Non-oil
Exports.International Formal of Academic Research in Accounting, Finance and
Management Sciences Vol. 5 No. 1 January.2015 PP190-198.www.hrmass.com
International Energy Agency (IEA), (2005). Energy Balances of OECD Countries and
Energy Balances of Non-OECD Countries (2003 Edition) Paris: IEA.
It O.K. (2010), The Impact of Oil Price Volatility on Macro-Economic activity in Russia.
Economic Analysis Working Paper 9(5): 1-10.
Ito, K. (2008). “Oil Price and the Russian Economy: A VEC Model Approach” Internal
Research Journal of Financial and Economics Issue 17.
Jennifer C. (2007) The Effct of Oil Prices on Exchange Rates. A Case Study of the Dominion
Republic Undergraduates Economic Review Volumes Issue 2 Article 46.
Jin, G. (2008). “The Impact of Oil Price Shock and Exchange Rate Volatility on Economic
Growth:- A Comparative Analysis for Russian, Japan and China” Research Journal of
International Studies – Issue 8.
John B. (2002). Dictionary of Economies Oxford University Press, Great Clerendom Street,
Oxford Ox2 6DP.
Jarque C. M. and Bera , A K.(1980); Efficient Tests For Normality, Homoscedasticity And
Serial Independence Of Regression Residuals. Economics Letters 6 (1980) 255-259
North-Holland Publishing Company
Kasman A. and Kasman A.(2005). Exchange Rate Uncertainty in Turkey and its Impacts on
Export Volume. METU Studies in Development 23, 41-58.
Kerlinger F.N (1973). Foundation of Behavioural Research (2nd Ed.). New York: Holt,
Rinehart and Winston.
Klein M. (1990). “Satoral effects of Exchange Rate fluctuations on United States Exports
Journal of International Money and Finance 9:299-308.
Kon S. J (1984). Models of Stock returns: A comparison J. Finance, 39: 147-165.
Kroner K.F and Lastrapes W.D (1991) “The Impact of Exchange Rate Volatility on
International Trade; Reduced Form Estimated Rising the GARCH – in-mean Model,
Journal of International Money and Finance 12,298-318
Klar B. Linder F. and maintains S.G (2012). Specification Tests for the Error Distribution in
GARCH models. Journal of Computational Statistics and Data Analysis 56:35873598.
Liiu H.C and Hung J.C (2010). Forecasting & P-100 Stock Index Volatility.The Role of
Volatility Asymmetry and Distribution Assumption in GARCH Models. Export
Systems with Application 37, 4928-4934.
Liu H.C., Lee Y.H and Lee M.C (2009).Forecasting China Stock Markets Volatility Via
GARCH Models under Skewed – GED Distribution Journal of Money, Investment
and Banking 7, 5-15.
IIARD – International Institute of Academic Research and Development
Page 47
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
Majidi M. (2006). “Impact of Oil on International Economy, International Economics
Course, Center for Science and Innovation Studies.
Mckemzie M.D (1998). The Impact of Exchange Volatility on Australian Trade Flows.
Journal of International Financial Markets, Institution and Money 8,21-38.
Mckenzie .M.D (1999). The Impact of Exchange Rate Volatility on German-Us Trade Flows
Journal of International Financial Markets, Institution and Money 7,73.
Mordi N.O. (2006). Challenges of Exchange Rate Volatility in Economic Management in
Nigeria. In the Dynamics of Exchange Rate in Nigeria, Central Bank pf Nigeria
Bullion, 30(3); 17-25.
Nikbakht L. (2009) Oil Prices and Exchange Rates; The case of Opec. Business Intelligent
Journal, 3(1) 83-92.
Narayem P., and Narayan, S. (2007). Modeling Oil Price Volatility.Energy Policy 35, 65496553.
Nworgu B.G (2006). Educational Research Basic Issues and Methodology.University Trust
Publishers, Nssuka.
Nkomo J.C (2006). Crude Oil Price Movements and their Impact on South Africa Journal of
Energy in Southern African 17(4): 55-32.
Nnena A.N (2010). Exchange Rate Fluation and Trade Flows in Nigeria; A Time Series
Economic Model.
Oladipupo A.O. and Ogheneove, O.F. (2011).Impact of Exchange Rate on balance of
Payment in Nigeria. African Research Reviews 5(4): 73-88.
Omisakin A.O. (2008). “Oil Price Shocks and the Nigerian Economy: A Forecast Error
Variance Decomposition Analysis” Journal of Economic Theory 2(4) 124 – 130.
Omjimite B.U and Akpolodje, G (2010).The Effect of Exchange Rate Volatility on the
Imports of ECOWAS Countries. Journal of Social Sciences, 4(2) : 3014-346.
Onudugo, V.A Ikpe, M. &Anowor.O.F (2013).Non-Linear Oil Export and Economic Growth
in Nigeria.A Time Series Economic model.International of Business Management &
Research.(IJBMR), 312-115-124.
Olowe R.A (2009). Modeling Niara/Dolla Exchange Rate Volatility: Application of GARCH
and Asymmetric Models. International Review of Business Research Papers 5:377398.
Onwumere J.U.J (2005). “Business and Economic Research Methods” Lagos: Don-Vinton
Limited.
Olugbenga F. and Kehinde O.S (2017), Modeling the Impact of Oil Price Volatility on
Investment Decision marking in marginal field’s Development in Nigeria. British
Journal of Economics, Management and Trade. 17 (1): 1-16, 2017; Articule No.
BJEMT.28175.
Oyetunji B. (2013) “Ojet Price and Exchange Rate Volatility in Nigeria Pg5-32
Ozoudo P.J. (2010) Reviving Nigeria’s Non-Oil Sector for Economic Development; CBN
Economic and Financial Review, 35, 4/12/2010.
Ozturk.I. and Kalyoney H. (2009).Exchange Rate Volatility and Trade.An Emperical
Investigation from Cross-Country Comparison.African Development Review 33,1241264.
Performance in Nigeria. Interdisciplinary Journal of Contemporary Research in Business,
4(1): 149-155.
Peterson K. (2000). An introduction to applied Econometrics; a Time Series Approach.
Qian Y. and Varaing’s P. (1992). “Does Exchange Rate Hinder Export?” International
Economies Department, World Bank working paper an International Trade, WPS 911.
Ranguindin M.E and Reyes R.G (2006). “The effects of Oil Price Shocks on the Philippine
Economy: A VAR Approach”. A Paper Presented at the University of the Philippines
IIARD – International Institute of Academic Research and Development
Page 48
International Journal of Applied Science and Mathematical Theory ISSN 2489-009X Vol. 3 No. 4 2017
www.iiardpub.org
School of Economies.
Rydbery T.H (2000). Realistic Statistical Modeling of Financial Data. International Statistical
Review 68(3) 233 – 258.
Samuel (2011) Empirical Investigation of Agriculture Export Trade in Nigeria (1975-2008).
A Case Study of Cocoa and Palm Kernel. Central Bank of Nigeria (CBN) Economy
Financial Review Volume 49/market, 2011.
Schnaley, G. (2007). Exchange Rate Volatility and Growth in Small Open Economics at the
Emu Periphery Working Paper Series No. 773.
SharmiriA. and Isa Z. (2009), modeling and Forecasting volatility of the malysian stock
markets. Journal of Mathematics and Statistics Vols. (3): 234-240, 2009.
Simkowitz M.A and Bleedless W.L. (1980). Asymmetric Stable distributed security returns.
J. Am. Stat. Assocj 75:306-312 http://www.Jstor.org/pss/2287449.
Spiro P.S (1990). Stock Market Overreaction to bad News in good times; A rational
expectation Equilibrium. Review of Financial Studies 12:975-1007.
Tatyana G. (2010) “Econometrics of Crude Oil Markets, Pg. 3-5
Taylor, S. (1986). Modeling Financial Time Series, Wiley, Chihester.
VeeDjGonpot P. and Sookia N. (2011). “Forecasting Volatility of USD/MUR Exchange Rate
using GARCH(1,1) model with GED and student’s-t Error” University of Mauritius
Research Journal – Volume 17-2011.
Wake Ford J. (2006) “Impact of Oil Price Shocks on the South African Macro economy;
History and Prospects in Alliterated and Shirred Growth in South Africa;
Determinants Constraints and Opportunities: Paper Prepared for the TIPS/DRRU
forum; Johannesburg, 18-20/10/2006).
Why some Firms Export Review of Economies and Statistics 86 (2), 561 – 569.
www.Investopedia.com/terms/c/Couter-oil.asp
www.Unctad.org/en/pub/pubframe.htm, 6th January, 2009.
Yeh Y. and Lee T. (2000). The International and Volatility asymmetry of unexpected Returns
in the during Stock Markets, Global Finance Journal II; 129 – 149.
Zhang Y. (2013). The links between the Price of Oil and the value of US Dollar.International
Journal of Energy Economies and Policy Vol. 3.www.econjournals.com.
IIARD – International Institute of Academic Research and Development
Page 49