Applied Energy 137 (2015) 110–116

Contents lists available at ScienceDirect

Applied Energy

journal homepage: www.elsevier .com/ locate/apenergy

Economic growth and biomass consumption nexus: Dynamic panelanalysis for Sub-Sahara African countries

http://dx.doi.org/10.1016/j.apenergy.2014.10.0170306-2619/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel./fax: +90 324 6514828.E-mail addresses: ilhanozturk@cag.edu.tr (I. Ozturk), fbilgili@erciyes.edu.tr

(F. Bilgili).

Ilhan Ozturk a,⇑, Faik Bilgili b

a Faculty of Economics and Administrative Sciences, Cag University, 33800 Mersin, Turkeyb Department of Economics, Faculty of Economics and Administrative Sciences, Erciyes University, Turkey

h i g h l i g h t s

�We examine long run dynamics of biomass energy consumption and GDP growth for 51 Sub-Sahara African countries.� The paper finds significant effect of biomass consumption on GDP in 51 Africa countries.� Economic growth is affected by biomass consumption, openness and population.

a r t i c l e i n f o

Article history:Received 7 June 2014Received in revised form 28 August 2014Accepted 4 October 2014

Keywords:GDP growthBiomass consumptionPanel data

a b s t r a c t

This paper investigates the long run dynamics of economic growth and biomass consumption nexus byapplying dynamic panel analyses for 51 Sub-Sahara African countries for 1980–2009 period. The resultsshow that economic growth is affected by biomass consumption, openness and population significantlyand positively in African countries. GDP elasticity with respect to biomass consumption is close to unityand the elasticities of GDP in terms of openness are found statistically significant (between 0.259 and0.348). According to homogeneous variance structure, one percent increase in variables of biomass,openness and population will lead GDP to increase by 1.818%, 0.269% and 0.676%, respectively. However,according to estimations from heterogeneous variance structure indicate that one percent increase inbiomass, openness and population variables will cause GDP to increase by 0.820%, 0.259% and 0.811%,respectively. In conclusion, this paper finds significant effect of biomass consumption on GDP in 51 Africacountries.

� 2014 Elsevier Ltd. All rights reserved.

Introduction

Energy plays a vital role for all countries in the world. Fluctua-tions in energy prices, energy dependency, environmentalproblems, climate change, energy security and limited fossil energysources have forced countries to replace fossil energy sources withrenewable and sustainable energy sources. Thus, renewableenergy, especially biomass has been accepted as a new energysource for sustainable development in the world. In addition,increases in the greenhouse gases (GHGs) have renewed concertedinterest in renewable energy to safeguard the environment tomitigate the impacts of climate change.

The majority of people in Sub-Sahara Africa (SSA) exceedinglydepend on biomass, especially combustibles for primary energy

generation for domestic cooking and heating purposes. Therefore,biomass is one of the prime sources of renewable energy in SSAwhich is used mostly by low income earners.

Some recent works emphasize (i) the negative impact ofbiomass usage on Green House Gas emission (GHG), (ii) and/orconsiderable cost of energy production from biomass, (iii) and/orpositive effect of biomass on GHG. Bilgili [1] following US datayields negative impact of biomass consumption on GHG and posi-tive influence of fuel usage on GHG. Khanna et al. [2] employingIllinois data emphasize the potential role of biofuels in mitigatingthe carbon emission. Acaroglu and Aydogan [3] considering Turk-ish data indicate both significant role of biofuels consumption inmitigating GHG and remarkable cost of biofuels production fromvegetable oils, animal fats and energy crops. Fischer et al. [4]observing EU data concludes that efficiency of biomass is subjectto change from country to country depending on land use effi-ciency and first or second generation feedstock and sustainabilityof feedstock. Reinhardt and von Falkenstein [5] explore that,

I. Ozturk, F. Bilgili / Applied Energy 137 (2015) 110–116 111

although biofuels have some negative effects on environment interms of GHG, the biofuels are considered favorable in comparisonwith their fossil alternatives.

This paper mainly considers the effect of biomass consumption,among other renewable consumption, on GDP growth in Sub-Sah-aran Africa since biomass is main primary energy source in thiscontinent. IEA [6] and Stecher et al. [7] reveal the shares of energysources within total primary energy supply in Sub-Saharan Africa.In terms of 2009, these energy sources and their shares in paren-theses are solid biofuels (61%), other renewables (0.2%), coal andpeat (19.7%), oil (14.1%), natural gas (2.7%), nuclear (0.6%) andhydro (1.4%), respectively [7]. Stecher et al. [7], hence, considerbiomass favorable and primary source of total energy supplyamong other energy sources in Sub-Sahara African countries.

Biomass energy, mainly through wood and charcoal, representsapproximatively 80% of the total energy consumption in Sub-Saharan Africa, and up to the equivalent of one third of the totalhousehold economy. Wood is the largest biomass energy resourceused today but biomass sources include also tree and grass cropsand forestry, agricultural, and urban wastes. Biomass is by far themain source of household fuel in Africa. In comparison, biomassaccounts for about 3% of final energy consumption in OECD coun-tries. In Sub-Saharan Africa, biomass is used by a large majority ofthe population to meet its energy needs, from a 68% in Kenya andup to a 94% in Burundi. Also, 80–90% of the residential energyneeds of low-income households are met by traditional biomass,mostly fuel wood or charcoal. Internationally, the total numberof people relying on traditional biomass as a source of heatingand cooking fuel went from an estimation of just under 2.4 billionpeople in 2002 and should, with 8% increase, reach 2.6 billion in2030. This figure is due to the important growth of biomass usein Africa, as the number is to increase from 646 to 996 million inthe same period (54% increase).1 In Africa, a total of 657 millionpeople (80% of the population) rely on the traditional use of biomassfor cooking. The IEA [8,9] estimates that the total final consumptionbiomass/waste share will lie between 51% and 57% by 2035 depend-ing on which scenarios are assumed. The total primary energydemand (TPED) for Africa is predominantly determined by biomassdemand with almost half of the energy demand being covered bybiomass and waste [9].

Growing interest for biomass energy is driven by the fact amongothers it contributes to poverty reduction in under developed anddeveloping countries and increase rural employment (biomassproduction is labor intensive). As also mentioned by Mohammedet al. [10], biomass consumption in SSA varies from country tocountry depending on the national access to electricity, resourcesavailability and renewable energy policy. The rate of biomass con-sumption in SSA cannot be likened to any other renewable energysources because biomass is considered the most easily acquiredenergy resource available for people from substandard economicbackgrounds. The study of Mohammed et al. [10] presented adetailed information about the status of renewable energy con-sumption and developmental challenges in Sub-Sahara Africa.

Eventually, the aim of this study is to estimate the relationshipbetween biomass energy consumption and economic growth in 51Sub-Sahara Africa countries for 1980–2009 period. The panel unitroot analyses, panel cointegration analyses, conventional OLS anddynamic OLS (DOLS) analyses are run throughout homogeneousand heterogeneous variance structures of the panel data to inves-tigate the relations between the variables.

This study is a complementary to the previous works and itdiffers from the existing literature of energy economics in many

1 http://ec.europa.eu/europeaid/where/acp/regional-cooperation/energy/documents/biomass_position_paper_en.pdf.

-

aspects. First, it is the first study in the literature that analyzes thecausal relationship between biomass energy consumption and eco-nomic growth for the analyzed countries. Second, it considers bothhomogeneous and heterogeneous variance structures for the panelestimations and by following dynamic OLS methodology as well asconventional OLS methodology for panel data. Third, it employedmultivariate model rather than bivariate by adding populationand openness variables into the model.

The rest of the paper is organized as follows: In the secondsection of the study, literature review is presented. Econometricmethodology and data are given in Section ‘‘Econometricmethodology and data’’. Section ‘‘Estimation results’’ consists ofthe empirical results, while the last section includes conclusionsand policy implications.

Literature review

The relation between energy consumption and economicgrowth in the context of causality has been investigated in manystudies during last two decade. An extensive literature survey ofthese studies can be seen in the study of Ozturk [11]. The empiricalresults of the studies which investigate the relationship betweenthese variables are sometimes inconsistent with each other dueto the using different data sets, alternative econometric methodol-ogies and different countries’ characteristics.

In recent years, the causal relationship between renewableenergy consumption and economic growth was investigated in somecountries and regions (see Table 1). But there is not any study relatedwith the Sub-Sahara Africa region that investigates relationshipbetween renewable energy consumption and economic growth.Thus, this paper will contribute the literature by filling this gap.

When we analyzed the previous studies shown in Table 1, wefound that in the most studies the bidirectional causality is rela-tionship is confirmed. In other words the feedback hypothesis isvalid in most cases. The feedback hypothesis is supported if thereexists bi-directional causality between energy consumption andeconomic growth. The meaning of the feedback hypothesis is thatthe energy consumption and GDP are jointly determined and affecteach other.

Econometric methodology and data

Econometric methodology

Econometric methodology follows panel unit root analysesunder the assumptions of common autoregressive (AR) and indi-vidual AR processes, considers panel common AR and individualAR cointegration analyses and employs panel Dynamic OrdinaryLeast Squares (DOLS) analyses under the both homogeneous andheterogeneous variance structures. Following Eq. (1) below, oneconcludes that panel data fyitg follows unit root process if jcijP 1.

yit ¼ ci þ ciyit�1 þ eit ; i ¼ 1;2; . . . ;N and t ¼ 1;2; . . . ; T: ð1Þ

An augmented form of Eq. (1) can be written as is in Eq. (2). Paneldata fyitg follows unit root process if qi ¼ ðci � 1Þ ¼ 0.

ð1� LÞyit ¼ ðyit � yit�1Þ ¼ qiyit�1 þuiwit þ eit ð2Þ

where L is lag operator and wit represents the variables of individualconstants and/or trends of ith section at time t. The common AR unitroot null hypothesis or within dimension indicates thatH0 : qi ¼ q ¼ 0, for all i. The individual AR null hypothesis orbetween dimension assumes that H0 : qi ¼ 0, for all i. When theprobability of null hypothesis falls into non-rejection area, onecan declare that panel data does not follow stationary process[12–15].

Table 1Literature survey on renewable energy consumption and economic growth.

Study Period Country Methodology Causality results

Sadorsky [31] 1994–2003 18 Emerging countries Panel model Bidirectional causalityPayne [32] 1949–2006 USA Toda-Yamamoto causality NeutralityApergis and Payne [33] 1985–2005 20 OECD countries Panel model Bidirectional causalityApergis and Payne [34] 1992–2007 Eurasia Panel model Bidirectional causalityMenyah and Rufael [35] 1960–2007 USA VAR model Unidirectional causality from growth to renewable

energyApergis and Payne [36] 1980–2006 6 Central American countries Panel model Bidirectional causalityApergis and Payne [37] 1990–2007 16 Emerging markets Panel model Unidirectional causality from economic growth to

renewable electricity consumption in the short-runand bidirectional causality in the long-run

Bildirici [38] 1980–2009 Developing and emerging countries ARDL model Bidirectional causality in the long runMenegaki [39] 1997–2007 Europe Panel model NeutralityApergis et al. [40] 1984–2007 Developed and developing countries Panel model Bidirectional causalityTugcu et al. [41] 1980–2009 G-7 countries ARDL model Bidirectional causalityBildirici [42] 1980–2009 10 Developing and emerging countries ARDL model Unidirectional causality from biomass energy to

growthAl-Mulali et al. [43] 1980–2009 108 Countries FMOLS model 79% Of countries bidirectional causality,

19% Neutrality causality,2% Of countries unidirectional causality from growthto renewable energy

Bildirici and Ozaksoy [44] 1960–2010 10 Countries ARDL model Bidirectional causality in the long runBildirici [45] 1990–2011 Transition countries ARDL model Bidirectional causality

112 I. Ozturk, F. Bilgili / Applied Energy 137 (2015) 110–116

The variable fyitg might follow stationary process at its level[I(0)] or it follows stationary process at its first difference [I(1)].The variables of [I(1)] might be considered in a long run equilib-rium relation if they are cointegrated. When they are not foundcointegrated, the related estimations would yield biased andinconsistent estimators. This paper, therefore, after running panelunit root tests, conducts panel cointegration tests between panelGDP variable and panel biomass consumption variable. One, how-ever, should be careful when he/she considers the cointegrationtests between GDP and biomass consumption, he/she needs toemploy some other possible related variables into GDP functionas well as biomass consumption variable in order for function tobe defined well. To this end, this paper establishes panel GDP func-tion for 51 African countries as below.

GDPit ¼ f ðPit ;Oit ;BitÞ ð3Þ

where Pit; Oit and Bit denote population of ith country at time t,openness of country i at time t and biomass consumption of ithcountry at time t, respectively. Then the long run equilibriumbetween GDPit and explanatory variables of Pit; Oit and Bit is shownbelow in (4).

GDPit ¼ @i þ #iPit þ niOit þ piBit þ eit ð4Þ

where eit denotes residuals from Eq. (4). One may conduct the panelcointegration tests through Eq. (5).

eit ¼ bieit�1 þ zit ð5Þ

The null hypothesis of no cointegration is H0 : bi ¼ 1 for all i.The alternative homogeneous and heterogeneous hypotheses areHA : bi ¼ b < 1 for all i and HA : bi < 1 for all i [16–19].

The rejection of null hypothesis of no cointegration yields theoutput that GDPit; Populationit; Opennessit and Biomassit have longrun relationship. One can estimate these long run equilibriumparameters by Ordinary Least Squares (OLS) and/or DynamicOrdinary Least Squares (DOLS). Following the Eq. (4), the panelmodel can be rewritten as below.

GDPit ¼ @i þ liXit þ eit ð6Þ

where Xit is m � n matrix of Populationit; Opennessit and Biomassit

and li is m � 1 vector of parameters. The correction of endogeneityand serial correlation problems from OLS regressions is possible by

DOLS regression through differenced leads and lags as is given in Eq.(7).

GDPit ¼ @i þ x0itlþXp

j¼�p

@ikDxitþp þ eit ð7Þ

Following Eq. (7), one can assert that if (a) both fGDPitg andfxitg are I(1) and if (b) they are cointegrated, then, long run param-eters of OLS and DOLS are estimated through methodologies of Kaoand Chiang. [20] and Pedroni [21].

l̂OLS ¼XN

i¼1

XT

t¼1

ðxit � �xiÞðxit � �xiÞ0 !�1 XN

i¼1

XT

t¼1

ðxit � �xiÞðyit � �xiÞ !

ð8Þ

l̂DOLS ¼ N�1XN

i¼1

XT

t¼1

pitp0it

!�1 XT

t¼1

pitðyit � �yiÞ !

ð9Þ

where yit and pit represent dependent variable of GDP and2ðpþ 1Þ � 1 vector of independent variables including ðxit � �xi;

Dxit�p; . . . ;DxitþpÞ, respectively.

Data

This paper follows panel data of 51 African countries as given inTable 2 for the period of 1980–2009. The variables are (i) GDP[tcgdp, Total PPP Converted GDP (real), G–K method, at currentprices (in milions $)], (ii) population [POP, Population (in thou-sands)], (iii) openness [openc, Openness at current prices (%)]and (iv) biomass consumption [used extraction of Biomass in kt].The first three variables are extracted from Penn World Table, ver-sion 8.0 [22] and the data for biomass is obtained from GlobalMaterial Flow Database [23].

One can have initial observation of natural logarithms of paneldata through Figs. 1–4. The figures give the line graphs together withtheir trends of panel variables of GDPit; Populationit; Opennessit

and Biomassit , respectively. These panel variables denote (i) naturallogarithm of mean of 51 African Countries’ GDP level, (ii) natural log-arithm of mean of 51 African Countries’ Population level, (iii) naturallogarithm of mean of 51 African Countries’ Openness index, and (iv)natural logarithm of mean of 51 African Countries’ Biomass

Table 2Panel African Countries.

1 Algeria 27 Malawi2 Angola 28 Mali3 Benin 29 Mauritania4 Botswana 30 Mauritius5 Burkina Faso 31 Morocco6 Burundi 32 Mozambique7 Cameroon 33 Namibia8 Cape Verde 34 Niger9 Central African Republic 35 Nigeria

10 Chad 36 Rep. Congo11 Comoros 37 Rwanda12 Cote d’Ivoire 38 Sao Tome and Principe13 Djibouti 39 Senegal14 DR. Congo 40 Seychelles15 Egypt 41 Sierra Leone16 Equatorial Guinea 42 Somalia17 Ethiopia 43 South Africa18 Gabon 44 Sudan19 Gambia 45 Swaziland20 Ghana 46 Tanzania21 Guinea 47 Togo22 Guinea-Bissau 48 Tunisia23 Kenya 49 Uganda24 Lesotho 50 Zambia25 Liberia 51 Zimbabwe26 Madagascar

y = 9.1361e0.0026x

R² = 0.9975

9

9.1

9.2

9.3

9.4

9.5

9.6

9.7

9.8

9.9

10

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Popu

la�o

n

Popula�on

Expon.(Popula�on)

Fig. 2. Mean of 51 African countries’ population.

y = 0.0005x2 - 0.0059x + 4.1821R² = 0.8873

4

4.05

4.1

4.15

4.2

4.25

4.3

4.35

4.4

4.45

4.5

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Ope

nnes

s

Openness

Poly. (Openness)

Fig. 3. Mean of 51 African countries’ openness.

y = 0.0001x2 + 0.0164x + 10.228R² = 0.997

10.3

10.4

10.5

10.6

10.7

10.8

10.9

Biom

ass C

onsu

mp�

on

I. Ozturk, F. Bilgili / Applied Energy 137 (2015) 110–116 113

consumption level, respectively. All panel variables except popula-tion are fitted well with their polynomials trends. The panel datafor GDPit; Opennessit and Biomassit follow the trends of [y =0.0009x2 + 0.0285x + 9.0819], [y = 0.0005x2 � 0.0059x + 4.1821]and [y = 0.0001x2 + 0.0164x + 10.228], respectively. The populationvariable is defined well by exponential trend of[y = 9.1361e0,0026x]. The trend lines of the variables, from Figs. 1–4,yield the goodness of fit measure R2 of [0.9932], [0.9975], [0.8873]and [0.9970], respectively. One may reach initial inspection throughfigures that all data are positively sloped.

From Figs. 1–4, then, one observes that the panel variables ofGDPit, Populationit, Opennessit, and Biomassit tend to go upwardannually from 1980 to 2009. The average slopes of GDPit, Popula-tionit, Opennessit, and Biomassit are 0.0574, 0.0247, 0.0092 and0.0205, respectively. One, considering mean slopes, hence, mayindicate that (i) GDPit variable follows the highest increasing ratethrough time among other variables, (ii) Populationit has the secondhighest upward trend among others, (iii) the slope of Biomassit var-iable follows closely that of Populationit and, finally, (iv) as GDPit

increases, all other variables increase as well. Although the outputfrom (iv) might indicate the existence of potential possible co-movements between GDP, Population, Openness and Biomass con-sumption, it does not necessarily mean that GDP has statisticallysignificant long run equilibrium with Population, Openness andBiomass consumption through time within 51 African countries.

y = 0.0009x2 + 0.0285x + 9.0819

R² = 0.9932

8.5

9

9.5

10

10.5

11

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

GDP

GDPPoly. (GDP)

Fig. 1. Mean of 51 African countries’ GDP.

10.2

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Biomass Consump�onPoly. (Biomass Consump�on)

Fig. 4. Mean of 51 African countries’ biomass consumption.

Whether or not the biomass consumption and other variables ofPopulation and Openness have statistically significant impact onGDP in 51 African countries will be determined next sectionthrough panel unit root and panel cointegration estimations.

Estimation results

Panel unit root tests results for the related variables are pre-sented in Table 3. The all within and between unit root tests for

Table 4African panel cointegration tests over the period 1980–2009.

Statistic Prob.

Within estimationsDF rho �163.304 0.000DF t rho �18.523 0.000DF rho star �260.458 0.000DF t rho star �14.245 0.000ADF �14.694 0.000t rho NT �923.714 0.000TN1 rho �358.827 0.000TN2 rho �358.710 0.000Standardized Panel v 47.636 0.000Standardized Panel rho �54.651 0.000Standardized Panel t nonparametric �15.608 0.000Standardized Panel t parametric �189.774 0.000

Between estimationsStandardized Group rho �6.503 0.000Standardized Group t nonparametric �6.267 0.000Standardized Group t parametric �4.147 0.000

114 I. Ozturk, F. Bilgili / Applied Energy 137 (2015) 110–116

natural log of GDP indicate that it is not stationary at its level andthat it is stationary, however, at its first differenced level accordingto within dimension test of Levin et al. [14]c and betweendimension tests of IPS W [15], ADF-Fisher Chi-square [24], PP-Fisher Chi-square [24], ADF-Fisher Chi-square [24]b and PP-FisherChi-square [24]b.

The population variable is found not stationary at its level byHadri [13] through within dimension methodology. All other testsfind population stationary at its level. The population variableagain has unit root at its level according to ADF-Fisher Chi-square[24]b and PP-Fisher Chi-square [24]b throughout betweendimension. The population variable, on the other hand, is foundstationary at its first differences by Levin et al. [14]c, ADF-FisherChi-square [24]b and PP-Fisher Chi-square [24]b. The opennessdata is found non stationary by Hadri [13]a – within dimensionand ADF-Fisher Chi-square [24]b – between dimension and PP-Fisher Chi-square [24]b – between dimension. One observes fromTable 3 that, panel data for openness follows stationary path atits first differences by all statistics at 1% significance level.

The biomass data has unit root according the tests of Hadri [13]a

and Breitung [12] at its level and it is considered nonstationaryeither at its level by ADF-Fisher Chi-square [24]b and PP-FisherChi-square [24]b. One also can claim that, from Table 3, the paneldata for biomass has stationary path at its first differences by all

Table 3African panel unit root tests over the period 1980–2009.

GDPit P

Test for levelsWithin dimensionLevin et al. [14] �0.6461

(0.259)Hadri [13]a 16.442

(0.000)Breitung [12] 13.202

(1.000)

Between dimensionIPS W [15] 0.9078

(0.818)ADF-Fisher Chi-square [24] 62.252

(0.999)ADF-Fisher Chi-square [24]b 4.670

(1.000)PP-Fisher Chi-square [24] 12.522

(1.000)PP-Fisher Chi-square [24]b 0.175

(1.000)

Tests for first differencesWithin dimensionLevin et al. [14] 3.113

(0.999)Levin et al. [14]c �3.768

(0.001)Breitung [12] 3.934

(1.000)

Between dimensionIPS W [15] �4.1317

(0.000)ADF-Fisher Chi-square [24] 139.969

(0.007ADF-Fisher Chi-square [24]b 396.824

(0.000)PP-Fisher Chi-square [24] 173.756

0.000)PP-Fisher Chi-square [24]b 567.311 1

(0.000)

a Follows null of stationarity. All other tests in the table are conducted through assumb Yields unit root tests of no intercept and no trend. All other tests, otherwise indicatc Yields unit root tests of individual intercept. All other tests in the table are, otherwi

statistics at 1% significance level. Overall, all variables here inAfrican panel work are considered I(1).

Table 4 provides panel cointegration analyses’ results. All statis-tics, within estimations (homogeneous variance structure) andbetween estimations (heterogeneous variance structure) yield the

opulationit Opennessit Biomassit

�28.996 �14.587 �15.484(0.000) (0.000) (0.000)21.7000 10.729 11.928(0.000) (0.000) (0.000)�5.2600 �6.639 1.176(0.000) (0.000) (0.880)

�14.9365 �12.598 �11.326(0.000) (0.000) (0.000)

391.768 330.288 298.058(0.000) (0.000) (0.000)77.955 46.488 5.241(0.963) (1.000) (1.000)

130.783 151.371 481.998(0.028) (0.0011) (0.000)2.1E�06 39.914 0.228

(1.000) (1.000) (1.000)

2.034 �24.116 �15.649(0.979) (0.000) (0.000)�11.368 �26.672 �22.437

(0.000) (0.000) (0.000)4.676 �20.059 �4.907

(1.000) (0.000) (0.000)

3.153 �17.371 �27.981(0.999) (0.000) (0.000)39.027 457.564 762.960(1.000) (0.000) (0.000)

496.036 939.455 1158.76(0.000) (0.000) (0.000)7.072 254.555 13432.9

(1.000) (0.000) (0.000)202.39 689.779 1942.66

(0.000) (0.000) (0.000)

ption of nonstationarity.ed, in the table are launched through individual intercept and trend model.se indicated, conducted through individual intercept and trend model.

Table 5Long run dynamics of GDPit in terms of Opennessit ; Populationit and Biomassit for African panel data over the period 1980–2009.

Homogeneous OLS Homogeneous adjusted OLS Homogeneous DOLS Heterogeneous DOLS

Biomassit 1.168 1.175 1.818 0.82(23.632) (133.168) (205.501) (126800.207)

Opennessit 0.348 0.348 0.269 0.259(17.561) (112.355) (86.619) (1251.15)

Populationit 0.99 0.983 0.676 0.811(23.724) (129.626) (88.937) (395.484)

R2 0.8679 0.8679 0.908 0.669

Adj. R2 0.8676 0.8676 0.9073 0.6664

I. Ozturk, F. Bilgili / Applied Energy 137 (2015) 110–116 115

conclusion that variables have long run equilibrium. Therefore onecan state that GDPit; Populationit; Opennessit and Biomassit havecointegration relationship.

All cointegration tests are done through (i) Gauss code of Kaoand Chiang [20], (ii) NPT 1.3 Program written by Chiang and Kao[25] and (iii) Pedroni’s RATS code [18]. Throughout within estima-tions, one can see the methodologies of DF and ADF in details inKao [16] and one may observe the methodologies of NT, TN1 andTN2 in Pedroni [17]. One also can find the details of within stan-dardized panel estimations in Pedroni [18]. As for the betweenestimations, one may find the between methodologies of standard-ized estimations in Pedroni [18,19]. Gauss code of Kao and Chiang.[20] and NPT 1.3 Program written by Chiang and Kao [25] withsome modifications are launched to obtain the parameter estima-tions. The t ratios are given in parentheses. Adjusted OLS estimatesare obtained through heteroskedastic-consistent standard errors.

Table 5 yields long run parameter estimates of GDP equation interms of biomass, openness and population. The estimation meth-odologies employ homogeneous estimations of OLS, homogeneousadjusted OLS, the homogeneous estimations of DOLS with one leadand two lags and the heterogeneous estimations of DOLS with onelead and two lags. All regressors are found statistically significantand they all have positive effects on GDP. The methodologies ofhomogeneous estimations of OLS, homogeneous adjusted OLS,the homogeneous estimations of DOLS with one lead and two lagsindicate that the panel GDP elasticity with respect to biomass con-sumption is elastic (estimated elasticity parameter is greater than1). The heterogeneous estimation of DOLS with one lead and twolags implies that GDP elasticity with respect to biomass consump-tion is close to unity. The homogeneous estimations of OLS, homo-geneous adjusted OLS, and the heterogeneous estimations of DOLSwith one lead and two lags conclude also that GDP elasticity withrespect to biomass consumption is close to unity. The elasticities ofGDP in terms of openness are found statistically significant as wellyet their magnitudes are within interval of 0.259–0.348. Accordingto homogeneous variance structure, one percent increase in vari-ables of biomass, openness and population will lead GDP toincrease by 1.818%, 0.269% and 0.676%, respectively. The estima-tions from heterogeneous variance structure, on the other handindicate that one percent increase in biomass, openness and popu-lation variables will cause GDP to increase by 0.820%, 0.259% and0.811%, respectively.

Among all estimation methodologies, one can consider mainlythe homogeneous DOLS estimations since they yield the highestR2 and adjusted R2 among others. Overall one may claim that thegrowth in biomass consumption effect on GDP growth in paneldata for 51 African countries.

Conclusions and policy implications

This paper examines possible cointegration relation betweenbiomass consumption and GDP growth. Accordingly this paper

seeks for long run relation of GDP with respect to independentvariables of biomass, population and openness in panel data for51 African countries from 1980 to 2009. To this end, the panel unitroot analyses, panel cointegration analyses, conventional OLS anddynamic OLS (DOLS) analyses are conducted considering bothhomogeneous and heterogeneous variance structures. The homo-geneous estimations of OLS, homogeneous adjusted OLS, thehomogeneous estimations of DOLS with one lead and two lagsand the heterogeneous estimations of DOLS with one lead andtwo lags yield the output that African panel GDP data is affectedby African panel data of biomass consumption, openness and pop-ulation significantly and positively.

The homogeneous estimations of OLS and DOLS conclude alsothat GDP elasticity with respect to biomass consumption is closeto unity. According to homogeneous variance structure, one per-cent increase in variables of biomass, openness and population willlead GDP to increase by 1.818%, 0.269% and 0.676%, respectively.The estimations from heterogeneous variance structure, on theother hand indicate that one percent increase in biomass, opennessand population variables will cause GDP to increase by 0.820%,0.259% and 0.811%, respectively.

Overall, by considering all tests conducted within the text, wefound a significant and substantive influence of biomass energyconsumption on economic growth. In other words, biomass energyconsumption plays a vital role in the economic growth processdirectly and/or as a complement to population and openness. Thus,energy conservation policies aimed at reducing energy consump-tion will have negative impacts on economic growth of Sub-SaharaAfrican countries.

As a policy implication, energy policies which improve the bio-mass energy infrastructure and biomass supply are the appropri-ate options for Sub-Sahara African countries since biomass energyconsumption has positive effect on economic growth. Therefore,Sub-Sahara African countries should invest more on renewableenergy sources and especially increase the usage of biomassenergy to have sustainable environment, reducing energy depen-dency and carbon dioxide emissions and promoting economicgrowth. In addition, demand side management (DSM) programs,which is the planning, implementation and monitoring of utilityactivities that are designed to influence customer use of electric-ity, should be implemented in these countries [26]. There are sev-eral studies examining the effects of DSM programs such asBergaentzlé et al. [27], Ardakani and Ardehali [28,29] and IEA[8]. According to the results of these studies, DSM is one of thetools that policy makers and energy players could promote inorder to: Increase energy and environmental efficiencies, improveexchanges and generation allocations between inter-connectedand increase system flexibility. Thus, these studies conclude thatDSM programs increase the energy efficieny. As reported by IEA[9], GDP has increased by 14.74% from 2000 to 2009, while elec-trical energy consumption (EEC) has had a growth of 2.38% forthe same period in USA. It is expected that decreasing EEC growth

116 I. Ozturk, F. Bilgili / Applied Energy 137 (2015) 110–116

rate for the USA is partially caused by the increase in implemen-tation of DSM programs.

Further research and development is necessary for biomasspotential to be fully appreciated and developed. As suggested byIRENA [30], the tremendous potential of energy crops can only berealized through an enormous increase in agricultural productionefficiency. Therefore, the expansion and improvement of currentagricultural production systems in Africa will play a crucial rolein the sustainable development of modern bioenergy systems.

References

[1] Bilgili F. The impact of biomass consumption on CO2 emissions: cointegrationanalyses with regime shifts. Renew Sustain Energy Rev 2012;16(7):5349–54.

[2] Khanna M, Önal H, Dhungana B, Wander M. Economics of herbaceousbioenergy crops for electricity generation: implications for greenhouse gasmitigation. Biomass Bioenergy 2011;35:1474–84.

[3] Acaroglu M, Aydogan H. Biofuels energy sources and future of biofuels energyin Turkey. Biomass Bioenergy 2012;36:69–76.

[4] Fischer G, Prieler S, van Velthuizen H, Lensink SM, Londo M, de Wit M. Biofuelproduction potentials in Europe: sustainable use of cultivated land andpastures. Part I: Land productivity potentials. Biomass Bioenergy2010;34:159–72.

[5] Reinhardt GA, von Falkenstein E. Environmental assessment of biofuels fortransport and the aspects of land use competition. Biomass Bioenergy2011;35:2315–22.

[6] IEA. Energy balance for Africa, 2009. Paris: OECD; 2009.[7] Stecher K, Brosowski A, Thrän D. Biomass potential in Africa. International

Renewable Energy Agency (IRENA); 2013.[8] IEA. Energy technology perspectives: scenarios and strategies for

2050. Paris: OECD; 2008.[9] IEA. World energy outlook 2010. Paris: IEA; 2010 (pp. 736).

[10] Mohammed YS, Mustafa MW, Bashir N. Status of renewable energyconsumption and developmental challenges in Sub-Sahara Africa. RenewSustain Energy Rev 2013;27:453–63.

[11] Ozturk I. A literature survey on energy-growth nexus. Energy Policy2010;38(1):340–9.

[12] Breitung J. The local power of some unit root tests for panel data. Adv Econom2000;15:161–78.

[13] Hadri K. Testing for stationarity in heterogeneous panel data. Econom J2000;3:148–61.

[14] Levin A, Lin C, Chu CJ. Unit root tests in panel data: asymptotic and finite-sample properties. J Econom 2002;108:1–24.

[15] Im KS, Pesaran MH, Shin Y. Testing for unit roots in heterogeneous panels. JEconom 2003;115:53–74.

[16] Kao C. Spurious regression and residual-based tests for cointegration in paneldata. J Econom 1999;90:1–44.

[17] Pedroni P. Panel cointegration: asymptotic and finite sample properties ofpooled time series tests with an application to the PPP hypothesis. IndianaUniversity, Department of Economics; 1995.

[18] Pedroni P. Critical values for cointegration tests in heterogeneous panels withmultiple regressors. Oxford Bull Econom Stat 1999;61:653–70.

[19] Pedroni P. Panel cointegration: asymptotic and finite sample properties ofpooled time series tests with an application to the PPP hypothesis. EconomTheory 2004;20:597–625.

[20] Kao C, Chiang MH. On the estimation and inference of a cointegratedregression in panel data, vol. 15. Nonstationarity Panels, Panel Cointegrationand Dynamic Panels, Elsevier Science Inc.; 2000. p. 179–222 [ISBN:0-7623-0688-2].

[21] Pedroni P. Purchasing power parity tests in cointegrated panels. Rev EconomStat 2001;83:727–31.

[22] [Penn World Table, version 8.0] Feenstra, Robert C, Robert Inklaar, TimmerMarcel P. The Next Generation of the Penn World Table; 2013.<www.ggdc.net/pwt>.

[23] [Global Material Flow Database] SERI, Dittrich M. Global material flowdatabase. 2012 version; 2012. <www.materialflows.net>.

[24] Maddala GS, Wu S. A comparative study of unit root tests with panel data andnew simple test. Oxford Bull Econom Stat 1999;61:631–52.

[25] Chiang MH, Kao C. Nonstationarity panel time series using NPT 1. 3 – a userguide. National Cheng-Kung University and Syracuse University; 2002.

[26] Gelazanskas L, Gamage KAA. Demand side management in smart grid: areview and proposals for future direction. Sustain Cities Soc 2014;11:22–30.

[27] Bergaentzlé C, Clastres C, Khalfallah H. Demand-side management andEuropean environmental and energy goals: an optimal complementaryapproach. Energy Policy 2014;67:858–69.

[28] Ardakani FJ, Ardehali MM. Long-term electrical energy consumptionforecasting for developing and developed economies based on differentoptimized models and historical data types. Energy 2014;65:452–61.

[29] Ardakani FJ, Ardehali MM. Novel effects of demand side management data onaccuracy of electrical energy consumption modeling and long-termforecasting. Energy Convers Manage 2014;78:745–52.

[30] IRENA. Biomass potentials in Africa. IRENA-DBFZ; 2013. <http://www.irena.org/DocumentDownloads/Publications/IRENA-DBFZ_Biomass%20Potential%20in%20Africa.pdf>.

[31] Sadorsky P. Renewable energy consumption and income in emergingeconomies. Energy Policy 2009;37:4021–8.

[32] Payne JE. On the dynamics of energy consumption and output in the US. ApplEnergy 2009;86(4):575–7.

[33] Apergis N, Payne JE. Renewable energy consumption and economic growth:evidence from a panel of OECD countries. Energy Policy 2010;38(1):656–60.

[34] Apergis N, Payne JE. Renewable energy consumption and growth in Eurasia.Energy Econom 2010;32:1392–7.

[35] Menyah K, Rufael YW. CO2 emissions, nuclear energy, renewable energy andeconomic growth in the US. Energy Policy 2010;38:2911–5.

[36] Apergis N, Payne JE. The renewable energy consumption-growth nexus inCentral America. Appl Energy 2011;88(1):343–7.

[37] Apergis N, Payne JE. Renewable and non-renewable electricity consumption-growth nexus: evidence from emerging market economies. Appl Energy2011;88(12):5226–30.

[38] Bildirici M. The relationship between economic growth and energyconsumption. J Renew Sustain Energy 2012;4(2):5.

[39] Menegaki AN. Growth and renewable energy in Europe: a random effect modelwith evidence for neutrality hypothesis. Energy Econom 2011;33:257–63.

[40] Apergis N, Payne JE, Menyah K, Rufael YW. On the causal dynamics betweenemissions, nuclear energy, renewable energy, and economic growth. EcolEconom 2012;69:2255–60.

[41] Tugcu CT, Ozturk I, Aslan A. Renewable and non-renewable energyconsumption and economic growth relationship revisited: evidence from G7countries. Energy Econom 2012;34:1942–50.

[42] Bildirici M. Economic growth and biomass energy. Biomass Bioenergy2013;50:19–24.

[43] Al-Mulali U, Fereidouni HG, Lee JY, Che Sab CNB. Examining the bi-directionallong run relationship between renewable energy consumption and GDPgrowth. Renew Sustain Energy Rev 2013;22:209–22.

[44] Bildirici M, Ozaksoy F. The relationship between economic growth andbiomass energy consumption in some European countries. J Renew SustainEnergy 2013;5(2).

[45] Bildirici M. Relationship between biomass energy and economic growth intransition countries: panel ARDL approach. GCB Bioenergy 2014. http://dx.doi.org/10.1111/gcbb.12092.