energy consumption and economic growth nexus in tanzania- an ardl bounds testing approach.pdf

7/24/2019 Energy consumption and economic growth nexus in Tanzania- An ARDL bounds testing approach.pdf

1/6

Energy consumption and economic growth nexus in Tanzania: An ARDL

bounds testing approach

Nicholas M. Odhiambo

Economics Department, University of South Africa (UNISA), P.O. Box 392, UNISA, 0003, Pretoria, South Africa

a r t i c l e i n f o

Article history:

Received 29 August 2008

Accepted 15 September 2008Available online 21 November 2008

Keywords:

Tanzania

Energy consumption

Economic growth

a b s t r a c t

In this paper, we examine the intertemporal causal relationship between energy consumption and

economic growth in Tanzania during the period of 19712006. Unlike the majority of the previous

studies, we employ the newly developed autoregressive distributed lag (ARDL)-bounds testing approach

by Pesaran et al. [2001. Bounds testing approaches to the analysis of level relationships. Journal of

Applied Econometrics 16, 289326] to examine this linkage. We also use two proxies of energy

consumption, namely total energy consumption per capita and electricity consumption per capita.

The results of the bounds test show that there is a stable long-run relationship between each of the

proxies of energy consumption and economic growth. The results of the causality test, on the other

hand, show that there is a unidirectional causal flow from total energy consumption to economic

growth and a prima-facie causal flow from electricity consumption to economic growth. Overall, the

study finds that energy consumption spurs economic growth in Tanzania.

& 2008 Elsevier Ltd. All rights reserved.

1. Introduction

Since 1970s, a number of studies have attempted to examine

the causal relationship between energy consumption and eco-

nomic growth in both developed and developing countries.

Unfortunately, the majority of the studies in developing countries

have concentrated mainly in Asia and Latin America, affording

sub-Saharan African (SSA) countries very little coverage and in

some instances none at all. In fact, empirical studies on countries

like Tanzania are almost non-existent. Even where such studies

have been undertaken, the empirical findings on the direction of

causality between energy consumption and economic growth

have been largely inconclusive. Over all, the empirical evidence

from previous studies on this subject shows that the causal

relationship between energy consumption and economic growthdiffers from country to country and over time. In addition,

previous studies have shown that the causality between the two

variables may be sensitive to the choice of the energy consump-

tion variable. Although the majority of the previous studies have

found a direct causal relationship between the various proxies

of energy consumption and economic growth, the literature

regarding the possible neutrality between energy consumption

and economic growth is growing in quantity and substance. The

majority of the previous studies on the causality between energy

consumption and economic growth have mainly used theresidual-based cointegration test associated with Engle and

Granger (1987) and the maximum likelihood test based on

Johansen (1988)and Johansen and Juselius (1990). Yet it is now

well known that these cointegration techniques may not be

appropriate when the sample size is too small (see Narayan and

Smyth, 2005). In addition, some previous studies have over-relied

on the cross-sectional data analysis, which generalises the causal

relationship between energy consumption and economic growth

across countries. The problem of using a cross-sectional method is

that by grouping together countries that are at different stages

of economic development, it fails to address the country-specific

effects of energy consumption on economic growth and vice

versa. In particular, the method fails to explicitly address the

potential biases induced by the existence of cross-countryheterogeneity, which may lead to inconsistent and misleading

estimates (see alsoOdhiambo, 2008;Ghirmay, 2004;Quah, 1993;

Casselli et al., 1996). It is against this backdrop that the current

study attempts to investigate the intertemporal causal relation-

ship between energy consumption and economic growth in

Tanzania using the autoregressive distributed lag (ARDL)-bounds

testing approach. The study uses two proxies of energy consump-

tion, namely the total energy consumption per capita and the

electricity consumption per capita against the real GDP per

capitaa proxy for economic growth. The rest of the paper is

structured as follows: Section 2 gives an overview of the energy

policies in Tanzania. Section 3 presents the literature review,

while Section 4 deals with the empirical model specification, the

ARTICLE IN PRESS

Contents lists available atScienceDirect

journal homepage:ww w.elsevier.com/locate/enpol

Energy Policy

0301-4215/$- see front matter & 2008 Elsevier Ltd. All rights reserved.doi:10.1016/j.enpol.2008.09.077

Tel.: +27 124294829.

E-mail address: [email protected]

Energy Policy 37 (2009) 617622
http://www.sciencedirect.com/science/journal/jepohttp://www.elsevier.com/locate/enpolhttp://localhost/var/www/apps/conversion/tmp/scratch_1/dx.doi.org/10.1016/j.enpol.2008.09.077mailto:[email protected]:[email protected]://localhost/var/www/apps/conversion/tmp/scratch_1/dx.doi.org/10.1016/j.enpol.2008.09.077http://www.elsevier.com/locate/enpolhttp://www.sciencedirect.com/science/journal/jepo


2/6

estimation technique and the empirical analysis of the regression

results. Section 5 concludes the study.

2. An overview of energy policies in Tanzania

The main sources of energy in Tanzania include biofuel, woodfuel, hydroelectric power, natural gas, biogas, coal reserves, wind

and solar energy. However, the most exploited energy source

is wood fuel, which is considered to be cheap and accessible to

the majority of Tanzanians who live mainly in rural areas. In fact,

the biomass energy resource, which consists of fuel-wood and

charcoal from natural forests and plantations, accounts for 90% of

the total energy consumption, while petroleum accounts for 8%

(SARDC, 2006). Petroleum and hydroelectric power on the other

hand, account for 8% and 1.2%, respectively. The remaining energy

sources, such as solar energy, wind, etc. account for only 0.8%. The

vast majority of Tanzanians do not have access to electricity. In

particular, the rural population is hard hit. While about 39% of the

urban population have access to electricity, only 2% of the rural

population have access to electricity. The government of Tanzaniahas implemented a number of policies which are aimed at

addressing the countrys energy needs. Recently, the government

decided to review the countrys energy policy of 1992, which

culminated in the National Energy Policy of 2003. The main aims

of this policy are to: (i) promote affordable and reliable energy

supplies throughout the whole country; (ii) reform the market for

energy services and establish an adequate institutional framework

to facilitate investment in the energy sector, taking into account

the environmental concerns in all energy activities; (iii) enhance

the development and utilisation of indigenous and renewable

(RE) energy sources; (iv) promote energy efficiency and conserva-

tion; and increase energy education and build gender-balanced

capacity in energy planning, implementation and monitoring

(Republic of Tanzania, 2005).

The consumption of energy in Tanzania has, however, shown a

mixed trend, especially since the 1990s. For example, the energy

consumption, measured in kilograms of oil equivalent per capita,

steadily decreased from 428.142 in 1980 to 352.062 in 1994,

before increasing to 358.427 in 1995. Although the consumption

later decreased to 353.185 and 347.882 in 1996 and 1997,

respectively, it later increased again to 360.055 in 1998. Since

1999 the energy consumption in Tanzania has been increasingconsistently, with the highest level of 498.29 since 1973 being

recorded in 2004.Fig. 1shows the trends of energy consumption

per capita and GDP per capita during the period 19942005 as

compared to 1980.

3. Literature review

The causal relationship between energy consumption and

economic growth has important implications from the theoretical,

empirical and policy standpoints. A unidirectional causality

running from electricity consumption to economic growth, for

example, implies that economic growth is dependent on energy

consumption, and a decrease in energy consumption may restraineconomic growth (see also Narayan and Singh, 2007, p. 1142).

A unidirectional causality running from GDP to energy consump-

tion, on the other hand, implies that a country is not entirely

dependent on energy for its economic growth, and that energy

conservation policies can be implemented with little or no

adverse effects on economic growth. Likewise, the finding of no

causality in either direction, i.e. the so-called neutrality hypoth-

esis, implies that energy conservation policies have no effect on

economic growth (seeAsafu-Adjaye, 2000;Paul and Bhattacharya,

2004).

On the empirical front, there exist four views regarding the

causal relationship between energy consumption and economic

growth. The first view argues that energy consumption Granger-

causes economic growth. This view has been widely supported by

ARTICLE IN PRESS

-5

0

5

10

15

20

25

30

1980

Year

Perc

ent

% Change in energy use per capita

% Change in GDP per capita

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005

Fig. 1. Trends of energy consumption per capita and economic growth during the period 19942005 as compared to 1980. Source: Authors own computation from the

World Development Indicators CD-ROM, 2007; International Financial Statistics Yearbook, 2007.

N.M. Odhiambo / Energy Policy 37 (2009) 617622618


3/6

Narayan and Smyth (2008)for the case of G7 countries;Narayan

and Prasad (2008)for the case Australia, Iceland, Italy, the Slovak

Republic, the Czech Republic, Korea, Portugal and the UK; Narayan

and Singh (2007)for the case of Fiji; Altinay and Karagol (2005)

for the case of Turkey; Wolde-Rufael (2004) for the case of

Shangai;Shiu and Lam (2004)for the case of China;Chang et al.

(2001)for the case of Taiwan; Yang (2000)for the case of Taiwan;

Cheng (1997)for the case of Brazil and Masih and Masih (1996)for the case of India, among others. The second view argues that it

is economic growth that drives the energy consumption in many

countries, and that as the economy grows the demand for energy

from different sections of the economy increases. The empirical

work, which is consistent with this view includes studies such as

Kraft and Kraft (1978)for the case of the USA,Cheng (1999)for the

case of India,Gosh (2002)for the case of India, Narayan (2005);

Narayan and Smyth (2005)for the case of Australia,Abosedra and

Baghestani (1989)for the case of the USA, Cheng and Lai (1997)

for the case of Taiwan, Masih and Masih (1997) for the case of

Indonesia,Hatemi-J and Irandoust (2005)for Sweden, Mozumder

and Marathe (2007) for the case of Bangladesh, among others.

The third view, however, maintains that both electricity con-

sumption and economic growth Granger cause each other, i.e.,that there is a bidirectional causality between electricity con-

sumption and economic growth. This view has been widely

supported by studies such asMasih and Masih (1997)for the case

of Korea and Taiwan, Yang (2000) for the case of Taiwan, Paul and

Bhattacharya (2004)for the case of India and Glasure (2002)for

the case of Korea. Finally, the fourth view maintains that there is

no causality between energy consumption and economic growth.

In other words, these studies assert that energy consumption

and economic growth are neutral with respect to each other.

This finding has been supported by the studies ofAkarca and Long

(1980)for the case of the USA,Yu and Hwang (1984)for the case

of the USA,Cheng (1995)for the case of the USA andCheng (1997)

for the case of Mexico and Venezuela, among others.

4. Estimation techniques and empirical analysis

4.1. CointegrationARDL bounds testing procedure

In this study the recently developed ARDL-bounds testing

approach is used to examine the long-run cointegration relation-

ship between each of the two proxies of energy consumption and

economic growth. The ARDL modelling approach was originally

introduced by Pesaran and Shin (1999) and later extended by

Pesaran et al. (2001). The ARDL cointegration approach has

numerous advantages in comparison with other cointegration

methods. Unlike other cointegration techniques, the ARDL does

not impose a restrictive assumption that all the variables understudy must be integrated of the same order. In other words, the

ARDL approach can be applied regardless of whether the under-

lying regressors are integrated of order one [I(1)], order zero [I(0)]

or fractionally integrated. Secondly, while other cointegration

techniques are sensitive to the size of the sample, the ARDL test is

suitable even if the sample size is small. Thirdly, the ARDL

technique generally provides unbiased estimates of the long-run

model and validt-statistics even when some of the regressors are

endogenous (see also Harris and Sollis, 2003). The ARDL model

used in this study can be expressed as follows:

Model 1total energy consumption and economic growth

DInyt a0 Xn

i1

a1iDInyti Xn

i0

a2iDInENCti a3Inyt1 a4InENCt1 mt

(1)

DInENCt b0 Xn

i1

b1iDInENCti Xn

i0

b2iDInyti b3Inyt1 b4IInENCt1 mt

(2)

Model 2electricity consumption and economic growth

DInyt f0 Xn

i1

f1iDInyti Xn

i0

f2iDInELECti f3Inyt1 f4InELECt1 mt

(3)

DInELECt d0 Xn

i1

d1iDInELECti Xn

i0

d2iDInyti d3Inyt1 d4InELECt1 mt

(4)

where Iny is the log of per capita real GDP; InENCthe log of total

energy consumption per capita; InELEC the log of electricity

consumption per capita; m the white noise error term and D thefirst difference operator. The per capita real GDP data was

computed from the International Financial Statistics Yearbook

(2007), while the total energy data and the electricity consump-

tion were obtained from the World Development Indicators

CD-ROM (2007).

The bounds testing procedure is based on the joint F-statistic

(or Wald statistic) for cointegration analysis. The asymptoticdistribution of the F-statistics is non-standard under the null

hypothesis of no cointegration between examined variables.

The null hypothesis of no cointegration among the variables in

Eq. (1) is (Ho: a3 a4 0) against the alternative hypothesis

(H1: a3aa4a0). In Eq. (2), the null hypothesis of no cointegra-

tion is (Ho: b3 b4 0) against the alternative hypothesis

(H1: b3ab4a0), while in Eq. (3), the null hypothesis of no

cointegration is (Ho:f3 f4 0) against the alternative hypoth-

esis (H1: f3af4a0). Finally, in Eq. (4), where the electricity

consumption per capita is the dependent variable, the null

hypothesis of no cointegration is (Ho: d3 d4 0) against the

alternative hypothesis (H1: d3ad4a0). Pesaran and Pesaran

(1997)and Pesaran et al. (2001)report two sets of critical values

for a given significance level. One set of critical values assumesthat all variables included in the ARDL model are I(0), while the

other is calculated on the assumption that the variables are I(1). If

the computed test statistic exceeds the upper critical bounds

value, then the Ho hypothesis is rejected. If the F-statistic falls into

the bounds then the cointegration test becomes inconclusive. If

theF-statistic is lower than the lower bounds value, then the null

hypothesis of no cointegration cannot be rejected.

4.2. Granger non-causality test

Once the long-run relationships have been identified in

Section 4.1, the next step is to examine the short-run and long-

run Granger causality between the two proxies of energy

consumption and economic growth. The traditional Grangersdefinition of causality is based on the notion that the future

cannot cause the past, but that the past can cause the future

(see also Takaendesa and Odhiambo, 2007). According to

Grangers definition of causality, a time series, Xt, causes another

time series, Yt, ifYt can be predicted better (in a mean-squared-

error sense) using past values ofXtthan by not doing so. That is if

past values ofXtsignificantly contribute to forecasting Yt, then Xtis said to Granger cause Yt. Causality from Y to X can also be

defined in the same way. That is, when past values of Ytsignificantly contribute to forecasting future values ofXt, then Ytis said to Granger cause Xt.

The Granger causality test method has been chosen in this

paper over other alternative techniques because of its favourable

response to both large and small samples. Guilkey and Salemi(1982), andGeweke et al. (1983), for example, have all shown that

ARTICLE IN PRESS

N.M. Odhiambo / Energy Policy 37 (2009) 617622 619


4/6

the Granger test outperforms the other methods in both large and

small samples. Other alternative test procedures that have been

suggested in the literature include:Sims (1972),Pierce and Haugh

(1977), and Geweke (1982), among others. The conventional

Granger causality test involves the testing of the null hypothesis

that Xt does not cause Yt and vice versa by simply running the

following two regressions

Yt a0 Xn

i1

a1iYti Xn

i1

b1iXti ut (5)

Xt b0 Xn

i1

a2iYti Xn

i1

b2iXti t (6)

whereut,etare the white noise error processes and n denotes thenumber of lagged variables.

The Null hypothesis that Xt does not Granger cause Yt is

rejected ifb1iare jointly significant (Granger, 1969). Likewise, the

null hypothesis that economic growth does not Granger cause

financial development is rejected if ais are jointly rejected.

However, the traditional causality tests suffer from two metho-

dological deficiencies (see also Odhiambo, 2004). First, these

standard tests do not examine the basic time series properties ofthe variables. If the variables are cointegrated, then these tests

incorporating different variables will be mis-specified unless the

lagged error-correction term is included (Granger, 1988). Second,

these tests turn the series stationary mechanically by differencing

the variables and consequently eliminate the long-run informa-

tion embodied in the original form of the variables. As opposed to

the conventional Granger causality method, the error-correction-

based causality test allows for the inclusion of the lagged error-

correction term derived from the cointegration equation. By

including the lagged error-correction term, the long-run informa-

tion lost through differencing is reintroduced in a statistically

acceptable way. The Granger causality model used in the current

study is based on the following model (see Odhiambo, 2007;

Narayan and Smyth, 2008).Model 1Causality test between total energy consumption

and economic growth

DInyt a0 Xn

i1

a1iDInyti Xn

i0

a2iDInENCti ECMt1 mt (7)

DInENCt b0 Xn

i1

b1iDInENCti Xn

i0

b2iDInyti ECMt1 mt

(8)

Model 2Causality test between electricity consumption and

economic growth

DInyt f

0X

n

i1

f1iDIny

tiX

n

i0

f2iDInELEC

t

i ECM

t

1 m

t

(9)

DInELECt d0 Xn

i1

d1iDInELECti Xn

i0

d2iDInyti ECMt1 mt

(10)

where ECMt1 is the lagged error-correction term obtained from

the long-run equilibrium relationship.

Although the existence of a long-run relationship between

[ENC, y] a n d [ELEc, y] suggests that there must be Granger

causality in at least one direction, it does not indicate the

direction of temporal causality between the variables. The

direction of the causality in this case can only be determined by

the F-statistic and the lagged error-correction term. While thet-statistic on the coefficient of the lagged error-correction term

represents the long-run causal relationship, the F-statistic on the

explanatory variables represents the short-run causal effect

(see Odhiambo, 2008; Narayan and Smyth, 2006). It should,

however, be noted that even though the error-correction term has

been incorporated in all the equations (7)(10), only equations

where the null hypothesis of no cointegration is rejected will be

estimated with an error-correction term (see also Narayan and

Smyth, 2006;Morley, 2006).

4.3. Empirical analysis

4.3.1. Stationarity test

Although the bounds test for cointegration does not require

that all variables be integrated of order 1 [ I(1)], it is important to

conduct the stationarity tests in order to ensure that the variables

are not integrated of order 2 [I(2)]. In fact, the F-test would be

spurious in the presence ofI(2) because both the critical values of

the F-statistics computed by Pesaran et al. (2001) and Narayan

(2005)are based on the assumption that the variables are I(0) or

I(1). The results of the stationarity tests on differenced variables

based on the PhillipsPerron and NgPerron tests are presented in

Tables 1 and 2.The results reported in Tables 1 and 2 show that after

differencing the variables once, all the variables were confirmed

to be stationary. The PhillipsPerron and NgPerron tests applied

to the first difference of the data series reject the null hypothesis

of non-stationarity for all the variables used in this study. It is,

therefore, worth concluding that all the variables used in this

study are not I(2).

4.3.2. Cointegration test

In this section, the long-run relationship between [ENC,y] and

[ELEC,y] is examined using the ARDL bounds testing procedure. In

the first step, the order of lags on the first differenced variables in

Eqs. (1)(4) is obtained from the unrestricted models by using an

Akaike Information Criterion and Schwartz Bayesian Criterion

(see the results of the above tests (not reported here)) show that

the optimal lag for both Models 1 and 2 is lag 2. Having

established the optimal lag length for Eqs. (1)(4), the next

step is to apply a boundsF-test to Eqs. (1)(4) in order to establish

ARTICLE IN PRESS

Table 1

Stationarity tests of variables on first differencePhilipsPerron (PP) test

Variable No trend Trend Stationarity status

DLy 10.1436*** 4.1167** Stationary

DLENC 3.6053** 6.2966*** Stati onary

DLELEC 6.0846*** 5.9431*** Stati onary

Note: (1) the truncation lag is based on Newey and West (1987)bandwidth.

(2) **and ***denote significance at 5% and 1%, respectively.

Table 2

Stationarity tests of variables on first differenceNgPerron test

Variable NgPerron test statistics (with trend) Stationarity status

MZ MZt MSB MPT

DLy 34.1774 3.94463 0.11542 3.68770 Stationary

DLENC 78.8378 6.26157 0.07942 1.22526 Stationary

DLELEC 66.7107 5.77029 0.08650 1.38849 Stationary

Asymptotic critical values(Ng and Perron, 2001,Table 1)

1% 23.8000 3.42000 0.14300 4.03000

5% 17.3000 2.91000 0.16800 5.48000

10% 14.2000 2.62000 0.18500 6.67000



5/6

a long-run relationship between the variables under study.

The results of the bounds test are reported inTable 3.

The results reported in Table 3show that when the real GDP (y)

per capita is used as the dependent variable in Model 1 and Model

2, the calculatedF-statistic is higher than the critical value at 10%

and 5% levels, respectively. However, when the total energy

consumption (ENC) in Model 1 and electricity consumption (ELEC)

in Model 2 are used as dependent variables, the calculatedF-statistics are lower than the lower-bound critical values at the

10% level. This implies that there is a unique cointegration vector

in Model 1 and also in Model 2.

4.3.3. Analysis of causality test based on error-correction model

Having found that there is a long-run relationship between

[y, ENC] in Model 1 and [y, ELEC] in Model 2, the next step is to

test for the causality between the variables used by incorporating

the lagged error-correction term into Eqs. (7) and (9), respectively.

The causality in this case is examined through the significance

of the coefficient of the lagged error-correction term and joint

significance of the lagged differences of the explanatory variables

using the Wald test. The results of these causality tests are

reported inTable 4.

The empirical results reported inTable 4show that there is a

distinct unidirectional causal flow from total energy consumption

to economic growth, both in the short run and in the long run

(see Model 1). The long-run causality from total energy con-

sumption to economic growth is supported by the coefficient

of the lagged error-correction term in the economic growth

function, which is negative and statistically significant, as

expected. The short-run causality from total energy consumption

to economic growth, on the other hand, is supported by the

F-statistic in the economic growth function, which is statistically

significant. The reverse causality from economic growth to total

energy consumption, however, is rejected by the lagged error-

correction term and the F-statistic in the energy function, which

are all statistically insignificant. The results also show that there is

a prima facie (short-run) causal flow from electricity consumption

to economic growth (see Model 2). The short-run causal flow from

electricity consumption to economic growth is supported by the

F-statistic in the economic growth function in the second model,which is statistically significant. The long-run causal flow from

electricity consumption to economic growth is, however, rejected

by the lagged error-correction term in the economic growth

function (Eq. (7)), which is statistically insignificant. A summary

of the causality test between the three variables is presented in

Table 5.

5. Conclusion

This study examines the intertemporal causal relationship

between energy consumption and economic growth in Tanzania

using two proxies of energy consumption, namely total energy

consumption and electricity consumption against per capita real

GDPa proxy for economic growth. The study attempts to answer

one fundamental question. Does energy consumption in Tanzania

spur economic growth? Although a number of studies have been

conducted on the causal relationship between energy consump-

tion and economic growth in a number of developing countries,

the majority of these studies have concentrated mainly on Asia

and Latin America, affording sub-Saharan African (SSA) countries

very little coverage and in some instances none at all. In addition,

previous studies on this subject suffer from two major weak-

nesses. Firstly, the majority of the previous studies have mainly

used the residual-based cointegration test associated with Engle

and Granger (1987)and the maximum likelihood test based on

Johansen (1988)andJohansen and Juselius (1990)to examine the

relationship between energy consumption and economic growth.Yet it is now well known that these cointegration techniques may

ARTICLE IN PRESS

Table 3

Bounds F-test for cointegration

Dependent variable Function F-test statistic

Model 1Total energy consumption and economic growth

y y(ENC) 3.799*

ENC ENC(y) 2.550

Model 2Electricity consumption and economic growth

y y(ELEC) 5.540**

ELEC ELEC(y) 1.026

Asymptotic critical values

1% 5% 10%

I(0) I(1) I(0) I(1) I(0) I(1)

6.027 6.760 4.090 4.663 3.303 3.797

Note: ** and * denote statistical significance at the 5% and 10% levels, respectively.

Asymptotic critical value bounds are obtained from Narayan (2005), p. 1987,

Appendix: Case II.

Table 4

Granger non-causality tests

Dependent variable Causal flow F-statistic t-Test on ECM R2

Model 1Total energy consumption and economic growth

Economic growth (y) Total energy consumption (ENC)-economic growth (y) 3.3674 (0.0208)** 2.044** 0.43

Total energy consumption (ENC) Economic growth (y)-total energy consumption (ENC) 9.3042 (0.0003)*** 0.77

Model 2Electricity consumption and economic growth

Economic growth (y) Electricity consumption (ELEC)-economic growth(y) 3.600 (0.0339)** 0.973 0.37

Electricity consumption (ELEC) Economic growth (y)-electricity consumption (ELEC) 0.95318 (0.4761) 0.24

** and *** denote statistical significance at 5% and 1% levels, respectively.

Table 5

Summary of causality tests

Variables Causality General conclusion

Economic growth

(DLy) and total

energy

consumption

(DLENC)

There is a distinct

unidirectional causal flow

from energy consumption to

economic growth.

Energy consumption

Grangercauses economic

growth.

Economic growth

(DLy) and

electricity

consumption(DLELEC)

There is a short-run

unidirectional causal flow

from electricity consumption

to economic growth.

Electricity consumption

Grangercauses economic

growth only in the short run.

N.M. Odhiambo / Energy Policy 37 (2009) 617622 621


6/6

not be appropriate when the sample size is too small (see Narayan

and Smyth, 2005). Secondly, some previous studies have over-

relied on the cross-sectional data analysis, which fails to address

the country-specific effects of energy consumption on economic

growth and vice versa. Using the ARDL-bounds test procedure, the

empirical results of this study show that there is a distinct

unidirectional causal flow from total energy consumption to

economic growth, both in the short run and in the long run. Theresults also show that there is a prima facie(short-run) causal flow

from electricity consumption to economic growth. Overall, the

study finds that energy consumption spurs economic growth in

Tanzania.

References

Abosedra, S., Baghestani, H., 1989. New evidence on the causal relationshipbetween United States energy consumption and gross national product.

Journal of Energy Development 14, 285292.Akarca, A.T., Long, T.V., 1980. On the relationship between energy and GNP: a

reexamination. Journal of Energy Development 5, 326331.Altinay, G., Karagol, E., 2005. Electricity consumption and economic growth:

evidence from Turkey. Energy Economics 27, 849856.Asafu-Adjaye, J., 2000. The relationship between energy consumption, energy

prices and economic growth: time series evidence from Asian developingcountries. Energy Economics 22, 615625.

Casselli, F., Esquivel, G., Lefort, F., 1996. Reopening the convergence debate: a newlook at cross-country growth empirics. Journal of Economic Growth 1 (3).

Chang, T., Fang, W., Wen, L., 2001. Energy consumption, employment, output, andtemporal causality: evidence from Taiwan based on cointegration and error-correction modeling techniques. Applied Economics 33, 10451056.

Cheng, B.S., 1995. An investigation of cointegration and causality between energyconsumption and economic growth. Journal of Energy and Development 21,7384.

Cheng, B.S., 1997. Energy consumption and economic growth in Brazil,Mexico and Venezuela: a time series analysis. Applied Economics Letters 4,671674.

Cheng, B.S., 1999. Causality between energy consumption and economic growth inIndia: an application of cointegration and error-correction modeling. IndianEconomic Review 34 (1), 3949.

Cheng, B.S., Lai, T.W., 1997. An investigation of cointegration and causality betweenenergy consumption and economic activity in Taiwan. Energy Economics 19,435444.

Engle, R.F., Granger, C.J., 1987. Cointegration and error-correction-representation,estimation and testing. Econometrica 55, 251278.

Geweke, J., 1982. Measurement of linear dependence and feedback betweenmultiple time series. Journal of the American Statistical Association 77,304313.

Geweke, J., Masee, R., Dent, W., 1983. Comparing alternative tests of causality intemporal systems: analytical results and experimental evidence. Journal ofEconometrics 21, 161194.

Ghirmay, T., 2004. Financial development and economic growth in sub-SaharanAfrican countries: evidence from time Series analysis. African DevelopmentReview 16, 415432.

Glasure, Y.U., 2002. Energy and national income in Korea: further evidence on therole of omitted variables. Energy Economics 24, 355365.

Gosh, S., 2002. Electricity consumption and economic growth in India. EnergyPolicy 30, 125129.

Granger, C.W., 1969. Investigating causal relations by econometric models andcross-spectral methods. Econometrica 37, 424438.

Granger, C.W., 1988. Some recent developments in a concept of causality. Journal of

Econometrics 39, 199211.Guilkey, D.K., Salemi, M., 1982. Small sample properties of three tests for Granger-causal ordering in a bivariate stochastic system. Review of Economics andStatistics 64, 668681.

Harris, R., Sollis, R., 2003. Applied Time Series Modelling and Forecasting. Wiley,West Sussex.

Hatemi-J, A., Irandoust, M., 2005. Energy consumption and economic growth inSweden: a leveraged bootstrap approach (19652000). International Journal ofApplied Econometrics and Quantitative Studies 24, 8798.

International Financial Statistics Yearbook, 2007. IMF, Washington, DCJohansen, S., 1988. Statistical analysis of cointegration vectors. Journal of Economic

Dynamics and Control 12, 231254.

Johansen, S., Juselius, K., 1990. Maximum likelihood estimation and inference oncointegration with applications to the demand for money. Oxford Bulletin ofEconomics and Statistics 52, 169210.

Kraft, J., Kraft, A., 1978. On the relationship between energy and GNP. Journal ofEnergy Development 3, 401403.

Masih, A.M., Masih, R., 1996. Energy consumption, real income and temporalcausality: resultsfrom a multi-country study based on cointegration and error-correction techniques. Energy Economics 18, 165183.

Masih, A.M.M., Masih, R., 1997. On the causal relationship between energy

consumption, real income prices: some new evidence from Asian NICs basedon multivariate cointegration/vector error correction approach. Journal ofPolicy Modeling 19, 417440.

Morley, B., 2006. Causality between economic growth and migration: an ARDLbounds testing approach. Economics Letters 90, 7276.

Mozumder, P., Marathe, A., 2007. causality relationship between electricityconsumption and GDP in Bangladesh. Energy Policy 35, 395402.

Narayan, P.K., 2005. The saving and investment nexus for China: evidence fromcointegration tests. Applied Economics 37, 19791990.

Narayan, P.K., Prasad, A., 2008. Electricity consumptionreal GDP causality nexus:evidence from a bootstrapped causality test for 30 OECD countries. EnergyPolicy 36, 910918.

Narayan, P.K., Singh, B., 2007. The electricity consumption and GDP nexus for FijiIslands. Energy Economics 29, 11411150.

Narayan, P.K., Smyth, R., 2005. Electricity consumption, employment and realincome in Australia: evidence from multivariate Granger causality tests.Energy Policy 33, 11091116.

Narayan, P.K., Smyth, R., 2006. Higher education, real income and real investmentin China: evidence from Granger causality tests. Education Economics 14,

107125.Narayan, P.K., Smyth, R., 2008. Energy consumption and real GDP in G7 countries:

new evidence from panel cointegration with structural breaks. EnergyEconomics 30, 23312341.

Newey, W.K., West, K.D., 1987. A simple, positive semi-definite, heteroskedasticityand autocorrelation consistent covariance matrix. Econometrica 55, 703708.

Ng, S., Perron, P., 2001. Lag length selection and the construction of unit root testswith good size and power. Econometrica 69, 15191554.

Odhiambo, N.M., 2004. Is financial development still a spur to economic growth? Acausal evidence from South Africa. Savings and Development 28, 4762.

Odhiambo, N.M., 2007. Supply-leading versus demand-following hypothesis:empirical evidence from three SSA countries. African Development Review19 (2), 257279.

Odhiambo, N.M., 2008. Financial depth, savings and economic growth in Kenya: adynamic causal linkage. Economic Modelling 25 (4), 704713.

Paul, S., Bhattacharya, R.B., 2004. Causality between energy consumption andeconomic growth in India: a note on conflicting results. Energy Economics 26,977983.

Pesaran, M., Pesaran, B., 1997. Working with Microfit 4.0: Interactive EconomicAnalysis. Oxford University Press, Oxford.

Pesaran, M., Shin, Y., 1999. An autoregressive distributed lag modeling approach tocointegration analysis. In: Strom, S. (Ed.), Econometrics and Economic Theoryin the 20th Century: The Ragnar Frisch centennial Symposium. CambridgeUniversity Press, Cambridge.

Pesaran, M., Shin, Y., Smith, R., 2001. Bounds testing approaches to the analysis oflevel relationships. Journal of Applied Econometrics 16, 289326.

Pierce, D., Haugh, L., 1977. Causality in temporal system: characterisation and asurvey. Journal of Econometrics 5, 265293.

Quah, D., 1993. Empirical cross-section dynamics in economic growth. EuropeanEconomic Review 37 (2-3).

Republic of Tanzania, 2005. Priority energy initiatives for Tanzania. A paperpresented by the Tanzanian Delegation at an EU Energy Initiative Workshop inMaputo (April).

SARDC, 2006. National Adaptation Programme of Action (NAPA) for Tanzania(April).

Shiu, A., Lam, P.L., 2004. Electricity consumption and economic growth in China.Energy Policy 32, 4754.

Sims, C., 1972. Money, income and causality. American Economic Review 62,540552.Takaendesa, P., Odhiambo, N.M., 2007. Financial development and economic

growth: an empirical analysis of two Southern African countries. Studies inEconomics and Econometrics 31 (3), 6180.

Wolde-Rufael, Y., 2004. Disaggregated energy consumption and GDP; theexperience of Shanghai, 195299. Energy Economics 26, 6975.

World Development Indicators CD-ROM, 2007. World Bank.Yang, H.Y., 2000. A note on the causal relationship between energy and GDP in

Taiwan. Energy Economics 22, 309317.Yu, E.S.H., Hwang, B.K., 1984. The relationship between energy and GNP: further

results. Energy Economics 6, 1861990.

ARTICLE IN PRESS


energy consumption and economic growth nexus in tanzania- an ardl bounds testing approach.pdf

Documents