The Relationship Between Commercial Energy Consumption
and Gross Domestic Income in Kenya
Susan M. Onuonga
The Journal of Developing Areas, Volume 46, Number 1, Spring 2012,
pp. 305-314 (Article)
Published by Tennessee State University College of Business
DOI: 10.1353/jda.2012.0022
For additional information about this article
http://muse.jhu.edu/journals/jda/summary/v046/46.1.onuonga.html
Access provided by Kenyatta University. Library (19 Apr 2013 02:49 GMT)
The Journal of Developing Areas
Volume 46
No. 1
Spring 2012
THE RELATIONSHIP BETWEEN
COMMERCIAL ENERGY CONSUMPTION AND
GROSS DOMESTIC INCOME IN KENYA
Susan M. Onuonga
Kenyatta University, Kenya
ABSTRACT
The causal relationship between economic growth and energy consumption represents a widely
studied topic in energy economics literature. Although it is well known that there is a strong
correlation between energy consumption and economic growth, the issue of causality is still
controversial. The paper investigated the causal relationship between energy consumption and
economic growth in Kenya using published data. By using the Ganger-causality Error Correction
Model, the results suggest that economic growth causes energy consumption in Kenya. The
implication of the study is that energy conservation measures would not lead to negative effects on
the country’s economic growth.
JEL Classifications: C32, C33, Q41 and Q43
Keywords: Commercial energy consumption, Gross Domestic Product, Granger- Causality Error
Correction model
Corresponding Author’s Email: E-mail: [email protected]
INTRODUCTION
Energy plays a key role in the economy from both the demand side and supply side. On
the demand side, energy is one of the basic items that a consumer buys to maximize her
utility. On the supply side energy is a key factor of production just as capital, labor, land
and materials, a fact that is hardly included in the macroeconomic growth theories.
As from the oil crisis of 1970s, and its effects on the economies has brought to
the right the importance of energy as a factor of production beside labor and capital.
Energy determines the economic growth of a country and its standards of living. There is
thus a causal relationship running from energy to economic growth or vice versa.
Although it is well known that there is a strong correlation between energy and economic
growth, the issue of causality is still controversial (Masih and masih, 1996; Stern, 1993;
Ebohon, 1996; Cheng and Lai ,1997; Ghali and El-Sakka, 2004).
This raises such important questions as: whether energy is a stimulus to growth,
or whether economic growth is a stimulus to energy consumption. The answers to these
questions have important implications for policy makers on energy.
If causality runs from energy to Gross Domestic Product (GDP) then, the
economy is energy dependent. According to Jumbe (2004), economic growth is
determined by energy consumption. If this is the case, a shortage of energy supply may
negatively affect economic growth leading to poor economic performance and increased
unemployment. On the other hand, if causality runs from GDP to energy, it means that
the country is not energy dependent. That is, energy conservation policies may be
implemented without adversely affecting economic growth and employment. In the event
306
that there is no causality in either direction, (referred to us as neutrality hypothesis), then
energy and GDP are not correlated. The country can implement energy conservation
policies without affecting economic performance of the country.
It is therefore important to empirically determine the relationship between
energy consumption and economic growth. This will add information on the current
debate on global warming where countries are supposed to reduce emissions of
Greenhouse Gases that pollute the environment. There will be added information on the
effect of conservation of energy on economic growth. If causality is from energy to GDP,
then conservation of energy will have a negative effect on economic growth. The results
are beneficial to energy investors and planners.
Kenya has developed Vision 2030, which contains a long-term strategy for
transforming the country into a middle-income economy in the next two decades. In the
first medium plan (2008-2012) of vision 2030, three projects dealing with energy are
identified, namely: a rapid rural electrification, an increase of energy access; and a least
cost power development programme to generate additional 505MW of electricity
(Republic of Kenya, 2007a). These projects are meant to enable the country achieve the
Vision. In this Vision, it is not clearly known whether energy drives the economy or vice
versa. Causality analysis of energy and economic growth is essential in the achievement
of this Vision.
LITERATURE REVIEW
Given the importance of energy- income causality analysis, many studies have been
carried out in both developed and developing counties. There is no consensus on the
results obtained. Kraft and Kraft (1978), in their study on USA economy, supported
unidirectional causality hypothesis for the period 1947-1974. Erol and Yu (1987)
investigated the relationship between energy consumption and GDP for England, Italy,
Germany, France, Japan and Canada using the 1952-1982 data. The results showed
bidirectional causality for Japan, unidirectional causality running from energy
consumption to GDP for Canada, and unidirectional causality running from GDP to
energy consumption for Italy and Germany. The study did not find any causality for
France and England. Stern (1993) examined the relationship between energy
consumption and GDP for USA using multivariate cointegration model. The results
showed no relationship between the two variables. This contradicts Kraft and Kraft
(1978) `s findings.
Ebohon (1996) investigated the causality between energy consumption and
economic growth for Tanzania and Nigeria. The study found a bilateral causality between
energy consumption and economic growth for both countries. Masih and Masih (1996),
in their study for the Philippines, found no causality between energy consumption and
economic growth. In another study, Masih and Masih (1997) found bidirectional causality
for Korea and Taiwan. In the study done by Chenge and Lai (1997) for Taiwan province
of China, found causality running from economic growth to energy consumption, and
from energy consumption to employment without feedback. Shyamal (2004) examined
the causal relationship between energy consumption and economic growth in India for
the period 1950-1996. Applying Engle-Granger cointegration and the standard Granger
307
causality methods, the study found bi-directional causality between energy consumption
and economic growth.
Some studies study the relationship between energy and economic growth by
separating energy into its components. Jumbe (2004), examined the relationship between
electricity consumption and GDP for Malawi for the period 1970-1999. The study found
bidirectional causality between electricity consumption and GDP. However, when the
study examined the relationship between electricity consumption and non-agricultural
GDP, the study found a unidirectional causality relationship between the two variables.
Ghosh (2002) examined the relationship between economic growth and electricity
consumption of India between 1950 and 1997. The study found a unidirectional causality
relationship from economic growth to electricity consumption.
From the literature review, it is observed that the results are mixed; including
bilateral, unidirectional and no relationships between energy consumption and economic
growth. It is therefore important to examine the case in Kenya, especially now that the
country has set up Vision 2030.
This paper addresses the issue of energy-income causality by using the Granger
causality error Correction model (ECM) approach. The following section discusses
energy profile in Kenya.
KENYA'S ENERGY PROFILE
In Kenya, sources of energy comprise both commercial and non-commercial alternatives.
Commercial sources of energy include petroleum products and electricity, while noncommercial ones, comprise biomass and to a lesser extent other alternative renewable
energy such as solar energy, wind power and biogas.
Crude oil is the leading source of the world modern energy supply. In Kenya,
petroleum fuel is a major source of modern energy accounting for about 20 per cent of
the total energy consumption. This form of energy is mainly used in the transport,
industrial and commercial sectors. Consumption of petroleum was 3121.8 thousand
tonnes in 2007, up from 3038.2 thousand tonnes in the previous year. The overall demand
is projected to rise by 2% per annum on average to reach 2.93 million tones by financial
year 2009/10 (Republic of Kenya, 2004). Petroleum explorations are still going on in
Kenya, and harvesting of this form of energy has not started. Therefore, the country relies
on imported petroleum for its local consumption.
Electric power is another important source of energy in Kenya. It contributes
about 10 per cent of the total energy consumed. This type of energy is mainly derived
from hydro, thermal oil and geothermal sources. The installed capacity by 2007 were
677.3, 389.3 and 128 Megawatts for hydro, thermal oil and geothermal, respectively. In
the past, the Kenya Power and Lighting Company (KPLC) carried out distribution and
sale of electricity. Reforms that were introduced in the power sub-sector between 1994
and 2000 led to the separation of power generation and marketing. Kenya Electricity
Generating Company Ltd (KENGEN) is now the major producer of electricity in Kenya.
KENGEN does not satisfy the demand for electricity in Kenya. Because of this, private
producers have been commissioned by KENGEN to supplement the power generated
especially in times of drought. These private producers are in two categories, namely: the
Independent Power Producers (IPPs) and the Emergency Power Producers (EPPs).
308
Most of the electricity in Kenya is hydro-based. The contribution of
hydroelectricity to total electricity supply was 55, 51, and 57 per cent in the years 2005,
2006 and 2007, respectively (Republic of Kenya, 2007b). The contributions of thermal oil
for the same periods were approximately 26, 30 and 27 percent, respectively. The
contribution of geo-thermal for the same years was 18, 18 and 17 percent of the total
electricity supplied, respectively. In addition to the above supplies of electricity, Kenya
has an agreement with the Uganda Electricity Board (UEB) to supply the country with 30
MW annually. Electric power, is consumed by all sectors, but mainly in the large and
medium commercial and industrial, domestic and small commercial and rural
electrification. The consumption of electricity has grown over the years and it is
expected to continue growing as the economy grows. and more especially, if the vision
2030 is to be achieved. Consumption of electricity by type of user in the year 2007 is
illustrated in the following pie chart.
FIGURE 1. ELECTRICITY CONSUMPTION
DSC
LM(C&I)
RE
Source of original data: Republic of Kenya, 2007, pp 186.
Notes: DSC, represents domestic and small commercial LM represents large and medium
(Commercial and Industrial), while RE stand for Rural electrification.
Biomass is another source of energy in Kenya. Biomass resources are derived
from forest formations such as closed forests, woodlands, bush lands, grasslands,
farmlands, plantations and agricultural and industrial residues. These resources include
wood fuels and agricultural residues. This type of energy leads in terms of total supply of
energy in the country, supplying about 70 per cent of Kenya’s total energy needs. About
80 per cent of the Kenyan population depends on wood fuel for domestic energy needs,
providing 93 per cent and 80 per cent of the rural household and urban area energy
requirements respectively (Republic of Kenya, 2000a). These forms of energy are mainly
utilized in the manufacturing commercial, transport and the residential sub-sectors.
For Kenya to meet its growing energy needs, the country faces both supply and
demand side management constraints. The government has come up with energy policy
to address these issues (Republic of Kenya, 2004). However, for such a policy there is
need to understand the causal relationship between energy consumption and GDP. The
purpose of this paper is to determine such a relationship. Because of lack of appropriate
309
official time series data on biomass energy, only electricity and oil are considered in this
study.
METHODOLOGY, DATA AND MODEL
In order to investigate the relationship between energy and Gross Domestic Product
(GDP) in Kenya, a two-step procedure was adopted. First, time series properties of the
data were investigated by use of unit root test and long- run relationship investigated by
use of cointegration analysis.
To carry out stationarity analysis, unit root test was used. Dickey and Fuller
(1981) approach was applied by use of the following model:
The idea of using many lagged difference terms is to minimize the autocorrelation in the
error term. In each case the Ho: δ=0 (there is a unit root). The test statistic on the
p
Yt  a0  a2 t  Yt 1   i  Yt i  U t
(1)
i 1
parameter  was thus computed and compared to the critical values of  in Dickey and
Fuller (1981). Failing to reject the null hypothesis implies that the series contain unit root
hence is non-stationary at levels.
In this study the ADF test was performed on log(GDP) and log(Energy) series.
In both equations, the constant and the linear trend were included since this represents the
most general specification.
Testing for co-integration was done using the Engle-Granger two-step procedure
(Granger, 1986; and Engle and Granger, 1987). The procedure involved testing whether
the regression residuals of the following long-run regressions were stationary:
log(percapita income) = a0+a1log(percapita energy)+ u1
(2)
log(percapita Energy) =b0+b1log(percapita income)+u2
(3)
where u1 and u2 are error terms assumed to be uncorrelated, with zero mean and constant
variance. The two equations were estimated using Ordinary Least Squares method (
OLS).
In the second step, the causal relation between energy and GDP of Kenya was
investigated. If the variables were non- stationary at levels and the linear combination of
them was non- stationary, the standard Granger causality test was used. If the series were
non- stationary at levels and there was a long- run relationship among the variables, then
the ECM approach was used (Yang, 2000).
The standard Granger causality analysis was however, done before the ECM.
Following Granger (1969), the statement that log(Energy) Granger causes log(GDP),
means how much of the current log(GDP) can be explained by past values of log(GDP)
and whether adding lagged values of log(Energy) can improve the explanation. The same
applies to the statement that log (GDP) Granger causes log (Energy). Granger causality,
however, does not mean that if one says log (Energy) Granger causes log(GDP), that
log(GDP) is the result of log(Energy). The standard Granger causality was investigated
by use of the following two regressions:
310
n
n
i 1
j 1
log(GDP) t   ai log(GDP) t i    j log(Energy) t  j  u1t
m
m
i 1
j 1
log(Energy) t   bi log(Energy) t i   j log(GDP) t  j  u2t
(4)
(5)
where u1t and u2t are error terms assumed to have zero means and uncorrelated, n and m
are lag lengths and should were set to be the longest period over which one series
predicted the other. The null hypotheses was that log(Energy) does not Granger cause
log(GDP) in regression 4 and that log(GDP) does not Granger cause log(Energy) in
regression 5. This was tested by F test for the joint significance of the parameters β j and
αj.
When using the ECM approach causality was tested by estimating the following
regressions:

n
n
Δlog(GDP)  a 0   i Δlog(GDP) t i   θ jΔlog(Energ y) t  j  π u1t 1  e1t (6)
i1
j1

m
m
 log(Energy)  a1   bi  log(Energy) t i   y j  log(GDP) t  j   u 2t 1  e2t (7)
i 1
j 1
Where Δ, is the first difference operator, while u^1t-1 and u^2t-1 are estimated
residuals from equations 2 and 3, respectively. In this approach log (Energy) Grangercauses log(GDP) if either the coefficients on lagged log(Energy) are statistically
significant or the coefficient on lagged error term (π) is statistically significant. Similarly,
in regression 7, log(GDP) Granger-causes log(Energy) if either the coefficients on
lagged log(GDP) are statistically significant or the coefficient on lagged error term (ρ) is
statistically significant. Four lags were used in both equations. The choice of lag length
was determined by testing the significance of the parameters θ,  , b and y. This lag
length was a reasonable period on which the independent variable can influence the
dependent variable.
The data used in this study were available from Kenya’s Economic Surveys and
Statistical Abstracts. From these documents, data on energy consumption per-capita and
Gross Domestic per-capita were extracted. Because of lack of data on other forms of
energy, only commercial energy (petroleum and electricity) were considered in this
paper. Energy consumption per-capita was calculated by dividing total energy consumed
each year by total population of the same year, while GDP per-capita was calculated by
dividing the total GDP of the country in each year by the corresponding year’s
population. All the data were converted into natural logs before causality analysis.
311
EMPIRICAL RESULTS
The empirical results are presented in three stages. First there is the stationarity test
,followed by cointegration results and lastly causality results are presented.
Stationarity Test
As discussed before it was necessary to test for stationarity of the series. The results are
presented in Table 1.
TABLE 1. ADF UNIT-ROOT TESTS FOR STATIONARITY
Variable
log(GDP)
log(E)
Level
-2.7(0.08)
-3.19(0.03)
First- Difference
-4.54 (0.000)
-3.43(0.003)
Notes: In brackets are Mackinnon p-values
The Mackinnon estimated values for rejection of the null hypothesis at 1%, 5%, and
10% are -4.25, -3.54, and -3.20. Therefore, the null hypothesis of non-stationarity of the
two series at levels was not rejected. The log of GDP per-capita and of energy
consumption per-capita were stationary at their first differences. Income and energy
series are therefore, integrated of order one. The lag length that removed serial correlation
for first difference of both series was unit, and these were determined by use of Akaike
Information Criterion (AIC) (Akaike, 1969).
Cointegration Test
The results for cointegration analysis are presented in Table 2.
TABLE 2. COINTEGRATION TEST
Regression
log(GDP)on log(Energy)
log(Energy) on log (GDP)
Notes: In brackets are Mackinnon p-values
ADF
-3.7(0.11)
-3.7(0.09)
The lag lengths as determined by AIC were 5 and 7 respectively for first and
second equations. The results for both equations showed that the residuals of both
estimated equations were stationary providing evidence that GDP and energy were
cointegrated. Therefore, the ECM approach was found to be appropriate in testing for
causality between energy consumption and GDP. Table 3, presents results for the
standard Granger causality test between energy consumption and GDP.
312
Causality Test
TABLE 3. GRANGER CAUSALITY TEST
Regression
log(GDP) on log(Energy)
F-value
0.71
log(Energy) on log(GDP)
13.41
As shown from table 3, only regression log(Energy) on log(GDP) is statistically
significant at 10% level suggesting that there is unidirectional causality running from
economic growth to energy consumption. What this means is that including past values of
log(GDP) in log(Energy) equation provides a better explanation of the current values of
log(Energy). In other words, economic growth causes energy consumption.
TABLE4. RESULTS OF ENGLE-GRANGER
CAUSALITY ERROR CORRECTION MODEL
∆log(GDP) on∆ log(Energy)
0.35(0.95)
0.013(0.7)
0.37
∆log(Energy) on∆ log(GDP)
4.4 (0.002)
-0.82(0.0016)
-3.61
The results in table 4 showed that there was unidirectional causality running
from GDP to energy consumption in Kenya over the sample period. This meant that
adding past values of Δlog(GDP) in the Δlog(energy) equation provided a better
explanation of changes in energy consumption. This agreed with the results obtained
earlier under the standard Granger causality test. Economic growth in Kenya leads to
increase in demand for energy.
CONCLUSIONS
This paper aimed at investigating causality linkage between energy consumption and
GDP in Kenya using secondary data over the period 1970-2005. According to EngleGranger cointegration procedure, there is long-run relationship between energy
consumption and GDP in Kenya. The study used the ECM approach but compared the
results with those of the standard Granger causality approach.
Results found, suggested that economic growth causes total energy consumption
in Kenya. The results were similar in both approaches. This result agreed with other
previous studies (Kraft and Kraft, 1978, Jumbe, 2004, Anjum, 2001) but disagreed with
others that found that energy drives the economy or a bidirectional causality or neutrality
hypothesis. One reason for this result is that Kenya’s economy is dominated by
agriculture as the major source of income, thus it is not yet an industrialized economy
where energy drives the economy. It is less dependent on energy as an input in its GDP
production as compared to advanced countries.
The result obtained in this study has important policy implications on Kenya`
energy and economic growth policy. Kenya uses a lot of foreign exchange to finance
fossil energy imports bill. The country’s import bill on crude petroleum and petroleum
products amounted to 29,014; 31,354; 40,788; 63,680; and 57,217 Kenya Shilling
millions in the years 1997, 1998, 1999, 2000 and 2001, respectively. Therefore, using
313
energy more efficiently and substituting oil for locally produced renewable energy
wherever possible could be a good policy. Energy conservation policy especially on
imported fossil energy cannot lead to any adverse side effects on economic growth in
Kenya. A lot of foreign exchange can be saved that can be utilized in other productive
areas leading to greater economic growth and other benefits such as increased
employment.
The other policy implication is that increased economic growth in Kenya, will
lead to increased demand for commercial energy holding other factors constant.
Therefore, there is need for the country to invest more on renewable energy that has more
advantages compared to fossil energy as far as environmental conservation is concerned.
In addition, given that there are some indications of petroleum potential within the
country, it will save a lot of money spent on importation of petroleum if the government
can provide conducive environment that can accelerate its exploitation.
REFERENCES
Akaike, H., “Fitting autoregressive models for prediction”. .Annals of the Institute of
Statistical Mathematics,1969, 21, PP.243-237.
Anjum , A. and Mohammad, S. B., “The relationship between energy consumption and
economic growth in Pakistan”. Asia-Pacific Development Journal , 2001,Vol 8 . PP.101-110.
Chenga, B.S and Lai, W.L., “An investigation of cointegration and causality between
energy consumption and economic activity in Taiwan” Energy Economics ,1997,vol.19. pp.435453.
Dickey and W.Fuller., “Likelihood ratio statistics for auto-regressive time series with
unit roots”. Econometrica,1981, vol.49,pp.1057-1072.
Ebohon, O.J ., “Energy, economic growth and causality in developing coutries: A case
study of Tanzania and Nigeria”. Energy policy, 1996, vol.24, pp.447-453.
Engle,R., and R.W.Granger.,“Co-integration and Error Correction: representation,
estimation, and testing”. Econometrica,1987,vol. 55, pp.251-276.
Erol, U. and E.S.H. Yu., “On the relationship between energy and income for
industrialized countries”. Journal of energy and Employment, 1987,vol.13, pp.113-122.
Ghali, K.H., and Eli-Sakka, M.I.T., “Energy use and output growth: A multivariate
cointegration analysis”.Energy Economics 2004, vol.26,pp. 225-238.
Ghosh,S., “Electricity consumption and economic growth in Taiwan”, Energy Policy.
2002, vol.28, pp. 125-129.
Glasure Y U and A.R Lee., “Cointegration, Error – Correction and the Relationship
between GDP and Energy: The case of South Korea and Singapore”, Resource and Energy
Economics ,1997,vol. 20, pp.17-25.
Granger, C. W. J., “Developments in the study of co-integrated economic variables”,
Oxford Bulletin of Economics and Statistics, 1986, vol. 48, no.3, pp.213-228.
Granger, C. W. J., “Investigating causal relations by econometric models and crossspectral methods”, Econometrica , 1969,vol. 37,no. 3, pp.424- 438.
Gujarati, D.N. , Basic Econometrics.1995, Tokyo: McGraw-Hill.
Gor, S., “Consumption of Commercial Energy in the Residential Sector of
Kenya.19701990”, Unpublished Master`s Thesis,1994, Kenyatta University, Nairobi.
Jumbe, C.B.L.,“Cointegration and causality between electricity consumption and GDP:
Empirical evidence from Malawi”. Energy Economics 2004,vol. 26, pp.61-68.
314
Kraft, J and Kraft, A ., “Relationship between energy and GDP”. Journal of Energy and
Development , 1978, vol .3, pp.401-403.
Masih,A.M.M and R. Masih. , “Energy Consumption, real income and temporal causality
from a mult-country study based on cointegration and Error-Correction modelling technique”.
Energy Economics, 1996,vol 18 issue 3, pp. 165- 183.
Mwakubo,S. etal., Strategies for securing energy Supply in Kenya, 2007,Discussion
paper No. 74, KIPPRA, Nairobi.
Nelson,C.R and C.I Plosser. . “Trends and random walks in macroeconomic time series”,
Journal of monetary economics, 1962, vol. 10 pp.139-162.
Onuonga, S.M ., “An Econometric Analysis of Energy Utilization in the Kenyan
Manufacturing sector”., 2008,Unpublished Ph.D Thesis, Kenyatta University, Nairobi.
Pindyck,R.S., and Rubinfeld,D.L., Econometric Models and Economic Forecasting.
1991.Singapore: McGraw-Hill Book Co.
Republic of Kenya. , Kenya Vision 2030: A Globally Competitive and Prosperous Kenya,
2007a,Nairobi: Government Printer.
Republic of Kenya ., Economic Survey. 2007b, Nairobi: Government Printer.
Republic of Kenya ., Statistics of Energy and Power, 1969-1977. Central Bureau of
Statistics, Ministry of Economic Planning and Community Affairs. 1978, Nairobi: Government
Printer.
Republic of Kenya.. Sessional Paper Number 4 on Energy, 1980,Nairobi: Government
Printer.
Republic of Kenya ., Study on Kenya's Energy Demand, Supply and Policy. Strategy for
Households, Small- Scale Industries and Service, Establishments.2001, Nairobi: Government
Printer.
Republic of Kenya ., Economic Survey. 2000a,Nairobi: Government Printer.
Republic of Kenya ., Development Plan. 2000b, Nairobi: Government Printer.
Republic of Kenya .(various Issues). Development Plan. Nairobi: Government Printer.
Republic of Kenya. (Various Issues). Statistical Abstract. Nairobi: Government Printer.
Republic of Kenya. , Draft National Energy Policy. 2004,Nairobi: Government Printer.
Stern ,D., “A Multivariate cointegration analysis of the role of Energy In the US Macroeconomy”, Energy Economics , 2000,vol . 22 (2) pp.267- 283.
Stern, D. “Energy and Economic Growth in USA, A Multivariater Approach”. Energy
Economics , 1993,vol. 15, pp.137-150.
Sasia, J.O. , Demand for Gasoline and Light Diesel in Kenya, 1987,M.A Thesis,
University of Nairobi.
Shiu, A and Lam, P.L., Electricity and Economic Growth in China, Energy Policy , 2004,
vol. 32, pp.47- 54.
Shyamal, P. , Energy economics, , 2004,vol .26, pp 977-983.
Sica, E ., “Causality between Energy and Economic growth: The Italian case”,
Department of Economics , 2006,University of Foggia, Foggia. Italy
Yang, H.Y. , “A note on the causal relationship between energy and GDP in Taiwan”.
Energy Economics , 2000, vol. 22 pp. 309-317.