output, renewable energy consumption and trade in africa

Output, renewable energy consumption and trade in Africa

Mohamed Safouane Ben Aïssa a, Mehdi Ben Jebli a,b, Slim Ben Youssef c,n

a LAREQUAD & FSEGT, University of Tunis El Manar, Tunisiab University of Jendouba, ISI du Kef, Tunisiac Manouba University, ESC de Tunis, LAREQUAD, Manouba 2010, Tunisia

H I G H L I G H T S

� We examine the relationship between renewable energy consumption, trade and output in African countries.� There is a bidirectional causality between output and trade in both the short and long-run.� In the short-run, there is no causality between renewable energy consumption and trade or output.� In the long-run, renewable energy consumption and trade have a statistically significant positive impact on output.� African authorities should encourage trade openness because of its positive impact on technology transfer and on output.

a r t i c l e i n f o

Article history:Received 7 May 2013Accepted 10 November 2013

Keywords:Renewable energy consumptionInternational tradeAfricaPanel cointegration

a b s t r a c t

We use panel cointegration techniques to examine the relationship between renewable energyconsumption, trade and output in a sample of 11 African countries covering the period 1980–2008.The results from panel error correction model reveal that there is evidence of a bidirectional causalitybetween output and exports and between output and imports in both the short and long-run. However,in the short-run, there is no evidence of causality between output and renewable energy consumptionand between trade (exports or imports) and renewable energy consumption. Also, in the long-run, thereis no causality running from output or trade to renewable energy. In the long-run, our estimations showthat renewable energy consumption and trade have a statistically significant and positive impact onoutput. Our energy policy recommendations are that national authorities should design appropriatefiscal incentives to encourage the use of renewable energies, create more regional economic integrationfor renewable energy technologies, and encourage trade openness because of its positive impact ontechnology transfer and on output.

& 2013 Elsevier Ltd. All rights reserved.

1. Introduction

The interaction between international trade and renewableenergy consumption has not been previously studied, and it isthe aim of the present paper which considers a panel of Africancountries. Nevertheless, it is accepted that the use of renewableenergy is linked to the transfer of technology which is directlylinked to international trade. It was recognized by both the Rio andJohannesburg conferences that trade helps achieving more effi-cient allocation of scarce resources, makes it easier for countries,rich and poor, to access environmental goods, services andtechnologies (World Trade Organization, 2011).

There are several empirical studies analyzing the causal rela-tionship between economic growth and the consumption ofrenewable energy (e.g. Apergis and Payne, 2010a, 2010b, 2011,2012; Sadorsky, 2009b). Other papers analyze the causal relation-ship between economic growth, renewable energy consumptionand CO2 emissions (e.g. Sadorsky, 2009a). All these studies provethat renewable energy consumption plays a vital role for increas-ing economic growth, and an energy policy planned to increasethe share of renewable energy in total energy consumption is veryeffective in reducing greenhouse gas emissions. Capital, labor, andrenewable energy consumption are not the only factors determin-ing economic growth. Indeed, there are other factors that can beincorporated in the production function to explain the growth ofgross domestic product (GDP) such as trade openness. This lattercan be defined as exports, or imports, or the sum of both dividedby the value of GDP.

Many papers study the relationship between energy consump-tion (total energy use), trade, and output. Lean and Smyth (2010a)

Contents lists available at ScienceDirect

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

Energy Policy

0301-4215/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.enpol.2013.11.023

n Corresponding author. Tel.: þ216 97363596.E-mail addresses: [email protected] (M.S. Ben Aïssa),

[email protected] (M. Ben Jebli), [email protected] (S. Ben Youssef).

Please cite this article as: Ben Aïssa, M.S., et al., Output, renewable energy consumption and trade in Africa. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.11.023i

Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

www.sciencedirect.com/science/journal/03014215

www.elsevier.com/locate/enpol

http://dx.doi.org/10.1016/j.enpol.2013.11.023



mailto:[email protected]








examine the dynamic relationship between economic growth,electricity generation, exports and prices for Malaysia. The resultsfrom Granger causality tests show the existence of a unidirectionalcausality running from economic growth to electricity generation.Lean and Smyth (2010b) examine the causal relationship betweenaggregate output, electricity consumption, exports, labor, andcapital in a multivariate model for Malaysia. They find that thereis a bidirectional causality between aggregate output and electri-city consumption. They conclude that Malaysia should adopt thedual strategy of increasing investment in electricity infrastructureand encouraging electricity conservation policies to reduce unne-cessary wastage of electricity. Narayan and Smyth (2009) find thesame conclusion for a panel of Middle East countries. Indeed, forthe panel as a whole, they find feedback effects between electricityconsumption, exports and GDP. Sadorsky (2011) uses panel coin-tegration techniques to show how trade can affect energy con-sumption for 8 Middle East countries. He finds a Granger causalityfrom exports to energy consumption, and a bidirectional relation-ship between imports and energy consumption in the short-run.In the long-run, he achieves that an increase in both exports andimports affect the demand of energy. Considering a sample of7 South American countries, Sadorsky (2012) confirms the long-run relationship between trade and energy consumption. Oneimportant consequence of these results is that environmentalpolicies designed to reduce energy use will reduce trade.

To our knowledge, there is no study, in any country and particularlyin Africa, trying to know the linkage between trade and renewableenergy consumption. The aim of this paper is to explore the causalrelationship between renewable energy consumption, trade, and out-put by considering a panel of 11 African countries.

The paper is structured as follows. Section 2 gives an idea aboutthe renewable energy sector and trade in Africa. Section 3 describesthe data. Section 4 is designated for descriptive statistics. Section 5deals with the empirical models and results, and Section 6concludes.

2. Renewable energy and trade in Africa

Many studies underline the great potential of Africa regardingrenewable energy production and consumption. Indeed, with theirsolar, wind, hydropower and geothermal capacities, among others,many African countries have set themselves ambitious strategicobjectives and launched large-scale integrated energy programsfrom which they expect benefits involving reduction of green-house gas emissions, direct and indirect job creation, local indus-trial development and the improvement of human capital.Renewable energies also offer the opportunity to serve isolatedregions remote from the national electricity grid and so improvethe access to energy particularly for the poorest.

According to the United Nations Industrial DevelopmentOrganization (2009), the most used renewable energy sourcesfor large-scale applications in Africa are hydropower, modernbiomass, geothermal, wind and solar. These sources are usuallygrid connected. Only about 5% of Africa's hydropower potentialestimated to 1750 TWh has been exploited. The total hydropowerpotential for Africa is equivalent to the total electricity consumedin France, Germany, United Kingdom and Italy put together. TheInga River in the Democratic Republic of Congo (DRC) holds greatpotential for hydropower generation in Africa with an estimatedpotential of around 40,000 MW. The DRC alone accounts for over50% of Africa's hydropower potential. Other countries with sig-nificant hydropower potential include Angola, Cameroon, Egypt,Ethiopia, Gabon, Madagascar, Mozambique, Niger and Zambia.Despite the low percentage use, large-scale hydropower so farprovides over 50% of total power supply for 23 countries in Africa.

The use of wind energy for a large-scale electricity productionhas been increasing faster than any other renewable energytechnology over the past decade. In 2007, new installations wereabout 21 GW, even more than hydropower. The development ofwind energy projects is primarily limited by the lack of preciseinformation about the wind potential. In terms of installedcapacity at the beginning of 2008, Africa had about 476 MW ofinstalled wind energy generation capacity compared to a globalestimation of 93,900 MW. Many countries as Morocco, Egypt,Tunisia, South Africa, and Ethiopia are developing large-scale windenergy projects.

Large-scale solar energy projects are very limited in Africabecause of their high cost. Many studies have established thatAfrica has great potential for concentrated solar thermal powergeneration from desert areas like the Sahara and Namibia. Egyptplans to install solar thermal plant of 300 MW by 2020. Severalcountries in North Africa are planning to install solar thermalplants in partnership with European countries. The United NationsEconomic Commission for Africa Office of North Africa (2012)reports a number of current initiatives such as the MediterraneanSolar Plan (MSP), the Euro-Mediterranean partnership, the agree-ments that exist between the European Union and some countriesof North Africa, the DESERTEC project. These partnerships aim todevelop projects, increase investments, produce and distributerenewable energies, strengthen interconnections and create anexpanding regional market for electricity.

Small-scale renewable energy systems are used to provide tocommunities energy services that are not accessible by existingconventional energy supply systems such as the electricity grid.Unfortunately, poor households have not benefited as much ashigh income households from solar photovoltaic (PV) systemsbecause of their relatively high costs.

Countries in Africa can increase their energy efficiency withoutdecreasing economic output or lowering the standards of living.Studies by the International Energy Agency show that in Africaenergy intensity, i.e., total energy consumed per GDP, is at leasttwice the world average. Experiences so far show that the adop-tion of energy efficiency is inhibited by barriers including lack ofappreciation of the benefits, initial capital requirements, resistanceto change, absence of policy and regulatory frameworks. Africa canincrease its energy efficiency by encouraging the use of renew-ables and more efficient technologies.

Recognizing that national energy markets are narrow (UnitedNations Industrial Development Organization, 2009), Africa is experi-encing a shift towards regionally integrated energy markets. RegionalEconomic Communities (RECs) as Economic Community of WestAfrican States (ECOWAS), East African Community (EAC) and SouthernAfrican Development Community (SADC) are already working onregionally integrated policy planning, development and energy accessprograms. These efforts should strengthen the use of renewableenergies. Indeed, RECs should play a more active role in promotingregionally integrated markets for renewable energy technologies thatare commercially viable in order to realize economies of scale thatattracts private sector investments. Moreover, RECs should encouragecoherence and greater networking among their member states topromote sharing of experiences and best practices in renewableenergy. This could be realized by establishing regional institutionsthat promote greater partnerships with similar institutions from otherregions of the world in order to promote research and technologytransfer, among other things.

3. Data

Annual data from 1980 to 2008 are collected for a sample of 11African countries, namely: Algeria, Comoros, Egypt, Gabon, Ghana,

M.S. Ben Aïssa et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎2






Kenya, Mauritius, Morocco, Sudan, Swaziland and Tunisia. Thecriterion of selection of countries is based on the availability ofdata and on the interest of empirical results. The multivariateframework for the analysis includes real gross domestic product(GDP, output) measured in constant 2000 US dollars. Renewableenergy consumption (REC) defined as total renewable electricityconsumption is measured in billions of kilowatt hours. It com-prises hydroelectric and total non-hydroelectric renewables (solar,wind, geothermal, tide and wave, biomass and waste). Exports(imports) are measured using merchandise exports (imports)measured in current US dollars and are converted to real valuesby dividing them by the price level of consumption (PC). Thecapital stock is measured by the gross fixed capital formation inconstant 2000 US dollars. Labor is measured as the total number oflabor force. Data on exports, imports, capital and labor areobtained from the World Bank (2010) World Development Indi-cators online data base. Data on renewable energy consumptionare obtained from the U.S. Energy Information Administration(2012). Data on PC are obtained from the Penn World Tablesversion 7.1 (Heston et al., 2012).

4. Descriptive statistics

Table 1 reports some summary statistics (Mean, Median,Maximum, and Minimum). The primary variables of interest inthis paper are output, renewable energy, exports and imports.Figs. 1–4 show the variation of each variable for the sample of 11African countries over the period 1980–2008.

Fig. 1 presents the evolution of real GDP (measured in constantbillion 2000 US dollars). Egypt has the biggest value of real GDPwith 145.59 billion of constant 2000 US dollars in 2008, whereasComoros has the smallest value with 0.13 billion of constant 2000US dollars in 1980. According to Fig. 1, we can see that Egypt takesthe first place, then Algeria, Morocco and Tunisia in the fourthplace. Comoros has the lowest level of real GDP.

Fig. 2 presents the evolution of the consumption of renewableenergy (measured in billion of kilowatt hours) and shows thatEgypt is the biggest consumer over all the period of observationwith 16.18 billion of kilowatt hours in 2007, and then we haveGhana and Kenya with 6.78 billion of kilowatt hours in 1997 and4.79 billion of kilowatt hours in 2007, respectively. The smallestconsumer of renewable energy is Comoros with 0.002 billion ofkilowatt hours consumed each year during the period 1980–2001.

Fig. 3 reports the variation of real merchandises exports(million US dollars) and shows that Algeria is the biggest inexports of merchandises with 1085.81 million US dollars in2008, and the smallest exporter is Comoros with 0.08 million USdollars in 1996.

Fig. 4 reports real merchandises imports (million US dollars)and shows that Egypt is the biggest in imports of merchandises

Table 1Summary statistics (output, capital, labor, renewable energy consumption, realexports, and real imports).Source: World Bank (2010) online database and Energy Information Administration(2012). Output and capital are measured in billion of constant 2000 US dollars.Labor force is measured in millions. Renewable energy consumption (REC) ismeasured in billion kilowatt hours. Real merchandise exports and imports aremeasured in million US dollars.

Description Output Capital Labor REC Exports Imports

Mean 19.71 4.45 6.29 2.12 73.60 93.67Median 7.59 1.51 5.59 0.66 35.50 42.51Maximum 145.59 34.90 26.31 16.18 1085.81 1139.72Minimum 0.13 0.02 0.13 0.002 0.08 0.35Cross sections 11 11 11 11 11 11

Fig. 1. Real GDP (billion 2000 US dollars).

Fig. 2. Renewable energy consumption (billion of kilowatt hours).

Fig. 3. Real merchandises exports (million US dollars).

M.S. Ben Aïssa et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 3






with 1139.72 million US dollars in 2008, whereas Comoros is thesmallest importer with 0.35 million US dollars in 1980.

5. Empirical models and results

Following Lean and Smyth (2010a, 2010b) and Sadorsky (2012),the relationship between economic growth, energy consumption,and trade is modeled using the production function. The models inLean and Smyth (2010a, 2010b) incorporate exports as tradevariable, whereas the model in Sadorsky (2012) incorporatesexports and imports in two separate specification models. Theaim of this paper is to investigate the relationship between output,renewable energy consumption and trade using the same speci-fication model as Sadorsky (2012). Output (Y) can be written as afunction of renewable energy consumption (REC), trade openness(O)1, capital (K), and labor (L):

Yit ¼ f ðRECit ;Oit ;Kit ; LitÞ ð1Þ

The natural logarithm of Eq. (1) gives the following equation:

yit ¼ αiþδitþβ1irecitþβ2ioitþβ3ikitþβ4ilitþεit ð2Þ

where i¼ 1;…;N for each country in the panel, t ¼ 1;…; T denotesthe time period and (ε) denotes the stochastic error term. Theparameters αi and δI allow for the possibility of country-specificfixed effects and deterministic trends, respectively.

To examine the relationship between renewable energy con-sumption and trade for a sample of 11 African countries, we usepanel cointegration techniques. These latter are interestingbecause models estimated from cross-sections of time series havemore freedom degrees and are more efficient than modelsestimated from individual time series. Panel cointegration techni-ques are particularly useful when the time series dimension ofeach cross-section is short. We begin our empirical analysis withstationarity tests, then panel unit root tests for cointegration,causality tests using Engle and Granger (1987), and we finish bythe long-run estimates.

5.1. Stationarity tests

Five panel unit root tests are performed to check for integrationand stationarity of our variables. For all panel units, the panelbased ADF test proposed by Levin et al. (2002) supposes homo-geneity in the dynamics of the autoregressive coefficients. How-ever, Maddala and Wu (1999) allow for as much heterogeneityacross units as possible by employing nonparametric methods inconducting panel unit root tests using the Fisher-ADF and Fisher-PP tests. The Carrion-i-Silvestre et al. (2005) test is a generalizationof the panel stationarity test of Hadri (2000). It is based on theassumption that the long-run variance is either homogeneous orheterogeneous. The null hypothesis of the Levin et al. (2002),Fisher-ADF, and Fisher-PP tests is a unit root and the alternativehypothesis is no unit root. The Carrion-i-Silvestre et al. (2005) testassumes stationarity under the null hypothesis.

Table 2 shows the result of the panel unit root tests and indicatesthat variables are not stationary at level, whereas at the firstdifference all of them are stationary at the 1% significance level.

5.2. Cointegration tests

Given that panel unit root tests reveal that all variables arestationary after first difference, we proceed testing for panelcointegration using Pedroni (2004) and Kao (1999) tests. To testthe existence of cointegration within a heterogeneous panel,Pedroni (2004) proposes two categories of cointegration testsand seven statistics. The first category is based on four statistics(panel statistics) including v-statistic, rho-statistic, PP-statistic andADF-statistic. These statistics are classified on the within-dimension and take into account common autoregressive coeffi-cients across countries. The second category is based on threestatistics (group statistics) including rho-statistic, PP-statistic andADF-statistic. These tests are classified on the between-dimensionand are based on individual autoregressive coefficients for eachcountry in the panel. The null hypothesis is that there is nocointegration, and the alternative hypothesis is that there iscointegration between variables. Panel cointegration tests ofPedroni (2004) are based on the residuals of Eq. (2).

Deviations from the long-run equilibrium relationship arerepresented by the estimated residuals, εit. The null hypothesisof no cointegration, ρi¼1, is tested via the following unit root teston the residuals:

εit ¼ ρiεit�1þωit ð3ÞThe results from the tests for the data set with exports are

reported in Table 3 and suggest that there are two panel statistics

Fig. 4. Real merchandises imports (million US dollars).

Table 2Panel unit root tests.

Variables LLC Fisher-ADF Fisher-PP CBL (HOM) CBL (HET)

y 0.63184 0.39371 0.31275 37.75a 26.13a

Δy �4.08180a 29.4177a 21.6182a 2.54 3.37rec �1.43362 8.73951 10.6198 13.08a 9.74a

Δrec �8.71745a 52.0488a 53.8316a 1.81 1.35k 3.71820 0.20746 0.17928 25.43a 14.70a

Δk �5.04686a 35.1207a 35.3387a 1.22 1.59l 1.35292 15.1002 17.8120 32.37a 14.24a

Δl �2.74208a 58.5439a 47.9929a 1.34 1.58ex 1.42444 0.54203 0.47253 27.49a 19.15a

Δex �6.32119a 51.3551a 51.6156a 4.86 5.12im 1.61855 0.49962 0.31792 42.82a 18.46a

Δim �8.17090a 52.4125a 52.9137a 3.05 1.83

Panel unit root test includes intercept and trend.a Critical values at the 1% significance level: LLC-0.84; Fisher-ADF 56.09; Fisher-

PP 61.15; CBL(HOM) 6.73; CBL(HET) 6.11.

1 Trade openness is incorporated into the production function by including realexports and real imports of merchandises in two separate specification models.Separate models are estimated for exports and imports because the export (ex) andimport (im) variables correlate highly (0.94).







(v-statistic and ADF statistic) of the within-dimension indicatingcointegration at the 5% and 10% significance levels, respectively.One group statistic of the between-dimension (group ADF-statis-tic) indicates cointegration at the 10% significance level. In total,the results from Table 3 reveal that there is at least some evidenceof cointegration between variables.

The results from the tests for the data set with imports arereported in Table 4 and suggest that there are two panel statistics(v-statistic and ADF statistic) of the within-dimension indicatingcointegration at the 10% and 5% significance levels, respectively.Two group statistics (PP-statistic and ADF-statistic) of thebetween-dimension indicate cointegration at the 10% and 5%percent significance levels, respectively. In total, four tests amongseven suggest the existence of long-run relationship betweenvariables.

It is useful to confirm the existence of cointegration for theerror correction model by using a second test for panel cointegra-tion proposed by Kao (1999), which is based on ADF statistic.

The result from Kao (1999) cointegration test for the data setwith exports reported in Table 5 indicates that we can reject thenull hypothesis of no cointegration at the 1% significance level.

It means that there is evidence of cointegration between variableswhen y (output) is defined as a dependent variable.

The result from Kao (1999) cointegration test for the data setwith imports reported in Table 6 indicates that we can reject thenull hypothesis of no cointegration at the 5% significance level. Itmeans that there is evidence of cointegration between variableswhen y (output) is defined as a dependent variable.

5.3. Causality tests

The finding of cointegration between variables confirms theexistence of short and long-run relationship between variablesand the error correction model corresponding to each model canbe estimated. To investigate the short-run dynamic and the long-run dynamic relationships between variables, Engle and Granger(1987) propose two-step procedure. The first step consists inestimating the long-run model specified in Eq. (2). The secondstep consists in defining the lagged residual obtained from Eq. (1)as the error correction term (ECT). The estimation of the dynamicvector error correction model is given as follows:

Δyit ¼ θ1iþ ∑q

j ¼ 1θ1;1ijΔyit� jþ ∑

q

j ¼ 1θ1;2ijΔrecit� j

þ ∑q

j ¼ 1θ1;3ijΔoit� jþ ∑

q

j ¼ 1θ1;4ijΔkit� j

þ ∑q

j ¼ 1θ1;5ijΔlit� jþλ1iECTit�1þμ1it ð4Þ

Δrecit ¼ θ2iþ ∑q


q

j ¼ 1θ2;2ijΔrecit� jþ ∑

q

j ¼ 1θ2;3ijΔoit� j

þ ∑q

j ¼ 1θ2;4ijΔkit� jþ ∑

q


Δoit ¼ θ3iþ ∑q


q


q


þ ∑q


q


Table 3Pedroni residual cointegration test results (y, rec, ex, k, l).

Alternative hypothesis: common AR coefs. (within-dimension)

Statistic Prob. Weighted Statistic Prob.

Panel v-Statistic 1.691271 0.0454nn 1.941998 0.0261nn

Panel rho-Statistic 1.614396 0.9468 1.417284 0.9218Panel PP-Statistic �0.128617 0.4488 �0.805810 0.2102Panel ADF-Statistic �0.346780 0.3644 �1.362670 0.0865n

Alternative hypothesis: individual AR coefs. (between-dimension)

Statistic Prob.

Group rho-Statistic 2.369551 0.9911Group PP-Statistic �0.936498 0.1745Group ADF-Statistic �1.553027 0.0602n

Null hypothesis: No cointegration.Trend assumption: Deterministic intercept and trend.Lag selection: Automatic SIC with a max lag of 5.

nn Critical values at the 5% significance level.n Critical values at the 10% significance level.

Table 4Pedroni residual cointegration test results (y, rec, im, k, l).

Alternative hypothesis: common AR coefs. (within-dimension)

Statistic Prob. Weighted Statistic Prob.

Panel v-Statistic 1.421514 0.0776n 1.623290 0.0523n

Panel rho-Statistic 1.694911 0.9550 1.677025 0.9532Panel PP-Statistic 0.050045 0.5200 �0.660107 0.2546Panel ADF-Statistic �0.387010 0.3494 �2.024870 0.0214nn

Alternative hypothesis: individual AR coefs. (between-dimension)

Statistic Prob.

Group rho-Statistic 2.568598 0.9949Group PP-Statistic �1.479789 0.0695n

Group ADF-Statistic �2.092685 0.0182nn

Null hypothesis: No cointegration.Trend assumption: Deterministic intercept and trend.Lag selection: Automatic SIC with a max lag of 5.

nn Critical values at the 5% significance level.n Critical values at the 10% significance level.

Table 5Kao cointegration test result (y, rec, ex, k, l).

t-Statistic Prob.

ADF �2.993029 0.0014nnn

Residual variance 0.001288HAC variance 0.001549

Null hypothesis: No cointegration.Trend assumption: No deterministic trend.Automatic lag selection based on SIC with max lag of 7.

nnn Critical values at the 1% significance level.

Table 6Kao cointegration test result (y, rec, im, k, l).

t-Statistic Prob.

ADF �2.145063 0.0160nn

Residual variance 0.001311HAC variance 0.001627

Null hypothesis: No cointegration.Trend assumption: No deterministic trend.Automatic lag selection based on SIC with max lag of 7.

nn Critical values at the 5% significance level.







Δkit ¼ θ4iþ ∑q


q


q


þ ∑q


q


Δlit ¼ θ5iþ ∑q


q


q


þ ∑q


q


ECTit ¼ yit� β̂1irecit� β̂2ioit� β̂3ikit� β̂4ilit ð9Þ

where Δ is the first difference operator; the autoregression laglength, q, is set at 2 and determined automatically by the SchwarzInformation Criterion (SIC); μ is a random error term; ECT is theerror correction term derived from the long-run relationship ofEq. (2). The significance of the error correction term and the short-run dynamics can be tested using t-statistic tests and Grangercausality F-statistic tests, respectively.

Table 7 reports short and long-run causality results of Grangertests for exports specific model and indicates that there isevidence of a bidirectional causality between output and exportsat the 10% level of significance in the short-run. However, there isno evidence of short-run causality between renewable energy andexports and between output and renewable energy. The errorcorrection term is statistically significant for output and exportsequations at the 1% significance level indicating that there isevidence of (i) a long-run causality from renewable energyconsumption, exports, capital and labor to output, and (ii) a

long-run causality from output, renewable energy consumption,capital and labor to exports.

Table 8 reports short and long-run causality results of Grangertests for imports specific model. In the short-run, there is evidenceof a bidirectional causality between output and imports at the 5%level of significance. However, there is no evidence of short-runcausality between renewable energy and imports and betweenoutput and renewable energy. The error correction term isstatistically significant for output and imports equations at the1% significance level indicating that there is evidence of (i) a long-run causality from renewable energy consumption, imports, capi-tal and labor to output, and (ii) a long-run causality from output,renewable energy consumption, capital and labor to imports.

Our study confirms the results of Sadorsky (2012) who hasevaluated the relationship between energy consumption (totalenergy use), trade, and output and has showed that there is abidirectional causality between exports (or imports) and output inthe short and long-run. It means that variations in exports (orimports) will affect economic growth and variations in economicgrowth will affect exports (or imports). This result is very inter-esting and shows that for these selected African countries anyreduction in trade openness, i.e., any reduction in either exports orimports, will be harmful for economic activity.

We show that there is no short-run causality between renewableenergy consumption and trade openness for the considered panel ofAfrican countries. Also, there is no causality from trade to renewableenergy consumption in the long-run. This means that any fluctua-tions in trade openness will not affect the consumption of renewableenergy. This result is not similar to that of Sadorsky (2012) who hasshowed that in the short and long-run there is evidence of a causalrelationship between exports and energy consumption.

Table 7Granger causality tests (model with exports).

Dependent variable Sources of causation (independent variables)

Short-run Long-run

Δy Δrec Δex Δk Δl ECT

Δy – 0.35791 (0.6994) 2.67662 (0.0704)n 1.46351 (0.2330) 0.04431 (0.9567) �0.040860 [�2.80796]nnn

Δrec 0.32509 (0.7227) – 1.33274 (0.2653) 0.57230 (0.5648) 0.27011 (0.7635) 0.006379 [0.89085]Δex 2.92869 (0.0549)n 0.23770 (0.7886) – 0.69971 (0.4975) 0.62006 (0.5386) �0.209228 [�3.56748]nnn

Δk 0.43421 (0.6482) 0.63494 (0.5306) 2.06257 (0.1289) – 0.16158 (0.8509) 0.071416 [1.48251]Δl 0.61034 (0.5438) 0.56512 (0.5689) 1.35445 (0.2596) 1.08832 (0.3381) – �0.003168 [�1.65966]

Lag lengths: 2.P-value listed in parentheses and t-statistic listed in brackets.

nnn Indicate statistical significance at the 1% level.n Indicate statistical significance at the 10% level.

Table 8Granger causality tests (model with imports).

Dependent variable Sources of causation (independent variables)

Short-run Long-run

Δy Δrec Δim Δk Δl ECT

Δy – 0.35791 (0.6994) 4.47685 (0.0121)nn 1.46351 (0.2330) 0.04431 (0.9567) �0.008055 [�2.98395]nnn

Δrec 0.32509 (0.7227) – 1.54622 (0.2147) 0.57230 (0.5648) 0.27011 (0.7635) 0.006933 [1.63208]Δim 3.32819 (0.0371)nn 0.62773 (0.5345) – 0.25045 (0.7786) 1.95937 (0.1427) �0.246238 [�3.84318]nnn

Δk 0.43421 (0.6482) 0.63494 (0.5306) 1.95271 (0.1436) – 0.16158 (0.8509) 0.066622 [1.71941]Δl 0.61034 (0.5438) 0.56512 (0.5689) 2.85333 (0.0592)n 1.08832 (0.3381) – 0.015146 [3.45757]

Lag lengths: 2.P-value listed in parentheses and t-statistic listed in brackets.

nnn Indicate statistical significance at the 1% level.nn Indicate statistical significance at the 5% level.n Indicate statistical significance at the 10% level.







In our study, there is no evidence of short-run causalitybetween renewable energy consumption and output, and thereis no long-run causality running from output to renewable energy.Thus, energy policy recommended to increase renewable energyconsumption will not affect the economic growth of Africancountries, in the short-run. This finding is not the same than thatof Apergis and Payne (2011, 2012) who found a bidirectionalrelationship between renewable energy consumption and eco-nomic growth in both the short and long-run.

5.4. Long-run estimates

The last step consists in the long-run estimation of Eq. (2) wherethe dependent variable is real GDP or output, and the independentvariables are renewable energy consumption, real exports (or imports),capital stock and labor force. In the context of a panel estimate, theordinary least squares (OLS) estimator is asymptotically biased and itsdistribution depends on nuisance parameters. To address this bias, weestimate the long-run structural coefficients of Eq. (2) by using the fullymodified OLS (FMOLS) and the dynamic OLS (DOLS) panel approachesproposed by Pedroni (2001, 2004). To correct for endogeneity andserial correlation, FMOLS uses a non-parametric approach, whereasDOLS uses a parametric approach. Since our variables are measured innatural logarithms, the coefficients estimated from the long-runcointegration relationship can be interpreted as long-run elasticities.

Table 9 reports the results for panel OLS, FMOLS and DOLS long-run estimates for Eq. (2) with exports. For the export, capital andlabor variables, the three approaches produce very similar results interms of sign, magnitude and statistical significance.2 Indeed, theirestimated coefficients are statistically significant at the 1% level andindicating a positive impact on output. The estimated coefficient ofrenewable energy consumption is not statistically significant underFMOLS and DOLS approaches, but is statistically significant at the1% level under the OLS approach with a positive impact on output.

For the FMOLS results, a 1% increase in exports increases output by0.19%, a 1% increase in capital increases output by 0.48%, and 1%increase in labor increases output by 0.24%. With OLS approach, a 1%increase in renewable energy consumption increases output by 0.03%.

Table 10 gives the results for panel OLS, FMOLS and DOLS long-runestimates for Eq. (2) with imports. For import, capital and laborvariables, the three approaches produce very similar results in termsof sign, magnitude and statistical significance. For these variables, theirestimated coefficients are statistically significant at the 1% level andindicate a positive impact on output. The estimated coefficient ofrenewable energy consumption denotes a positive impact on outputand is statistically significant at the 1%, 5% and 10% levels with OLS,FMOLS and DOLS approaches, respectively.

For the FMOLS results, a 1% increase in renewable energyconsumption increases output by 0.05%, a 1% increase in imports

increases output by 0.21%, a 1% increase in capital increases outputby 0.51%, and a 1% increase in labor increases output by 0.16%.

6. Conclusion

This paper explores the relationship between renewable energyconsumption, trade and output for 11 African countries during theperiod 1980–2008. Exploring renewable energy and trade in Africais interesting because many studies underline the great potentialof Africa regarding renewable energy production and consump-tion, and because the use of renewable energy is linked to thetransfer of technology which is directly linked to internationaltrade. The aim of this study is to determine whether internationaltrade in African countries has an impact on renewable energyconsumption. Our specific model is similar to that developed bySadorsky (2012) in which he estimates the impact of trade onenergy consumption in a sample of 7 South American countries.

The Granger causality test indicates that there is evidence of abidirectional causality between output and trade (exports or imports)in both the short and long-run. These empirical results mean thatinternational trade has a positive impact on the real GDP of the sampleof 11 African countries studied. They confirm as previous studies andinternational organizations' recommendations that international tradeis beneficial for developing countries because of, among other reasons,the gain of technology transfer through trade.

We show that there is no short-run causality between renewableenergy consumption and trade openness for the considered panel ofAfrican countries. Also, there is no causality from trade to renewableenergy consumption in the long-run. This means that trade opennesshas no direct effect, in both the short and long-term, on renewableenergy consumption. However, we think that an indirect effect mayexist in the short-term and more probably in the long-term, fromtrade to renewable energy through technology transfer. Indeed,international trade helps the transfer of technologies, but a relativelylong time is needed for African countries to build the necessaryhuman and physical capacities for producing renewable energies.

We prove that there is no evidence of short-run causality betweenrenewable energy consumption and output, and there is no long-runcausality running from output to renewable energy. One reason is thatthe renewable energy consumed in most African countries are weakcompared to the non-renewable energy consumed. The proportions ofthe renewable electricity consumed compared to the total electricityconsumed (in percentage) for the year 2008 and for the consideredpanel of countries are:3 Algeria (0.05), Comoros (0.84), Egypt (1.71),Gabon (6.33), Ghana (18.20), Kenya (9.10), Mauritius (2.89), Morocco(0.68), Sudan (3.23), Swaziland (3.68) and Tunisia (0.07). Indeed, fiscalincentives (emissions taxes, pollution permits, subsidies etc.) are notsufficiently important to induce producers to use renewable energy.

Table 9Panel OLS, FMOLS and DOLS long-run estimates (model with exports).

Variables rec ex k l

OLS 0.032927 (0.0051)nnn 0.195382 (0.0000)nnn 0.468869 (0.0000)nnn 0.244243 (0.0000)nnn

FMOLS 0.034133 (0.1736) 0.195467 (0.0001)nnn 0.479936 (0.0000)nnn 0.236151 (0.0000)nnn

DOLS 0.032927 (0.1985) 0.195382 (0.0001)nnn 0.468869 (0.0000)nnn 0.244243 (0.0000)nnn

Cointegrating equation deterministics: intercept and trend.All variables are measured in natural logarithms.P-values listed in parentheses.

nnn Critical values at the 1% significance level.

2 According to Banerjee (1999), the FMOLS and DOLS long-run estimates areasymptotically similar for more than 60 observations. In our panel data we have319 observations.

3 Data on non-renewable energy consumption are obtained from the U.S.Energy Information Administration (2012). Non-renewable energy consumption isthe total non-renewable electricity consumption produced using oil, natural gasand coal.







Estimated long-run elasticities show that renewable energyconsumption and trade (exports or imports) have a statisticallysignificant positive impact on real GDP. For the model withexports, a 1% increase in renewable energy consumption increasesoutput by 0.03%, and a 1% increase in exports increases output by0.19%. For the model with imports, a 1% increase in renewableenergy consumption increases output by 0.05%, and a 1% increasein imports increases output by 0.21%.

Our energy policy recommendations are the following. First,national authorities have to design appropriate fiscal incentives toencourage the use of renewable energies instead of non-renewable energies. Innovation subsidies are particularly interest-ing because abandoning a polluting technology for a greentechnology still needs important investments, especially in Africancountries. This will undoubtedly increase the proportion of renew-able energy compared to the total energy used in African coun-tries. Second, and as recommended by the United NationsIndustrial Development Organization (2009), regional economiccommunities (RECs), as the economic community of west Africanstates (ECOWAS), should strengthen the use of renewable energiesby promoting regionally integrated markets for renewable energytechnologies in order to realize economies of scale that attractprivate sector investments. RECs can establish regional institutionsthat promote greater partnerships with similar institutions fromother regions of the world in order to promote research andtechnology transfer. Third, more trade openness with its positiveimpact on technology transfer, can greatly help African countriesto diffuse the adoption of production technologies using renew-able energy, while increasing their output.

Acknowledgments

We would like to express our sincere gratitude to the editor, ananonymous referee and Rafik Baccouche whose constructivecomments and suggestions have considerably improved the earlierversion of this paper. All remaining errors are ours.

References

Apergis, N., Payne, J.E., 2010a. Renewable energy consumption and economicgrowth evidence from a panel of OECD countries. Energy Policy 38, 656–660.

Apergis, N., Payne, J.E., 2010b. Renewable energy consumption and growth inEurasia. Energy Econ. 32, 1392–1397.

Apergis, N., Payne, J.E., 2011. The renewable energy consumption–growth nexus inCentral America. Appl. Energy 88, 343–347.

Apergis, N., Payne, J.E., 2012. Renewable and non-renewable energy consumption-growth nexus: evidence from a panel error correction model. Energy Econ. 34,733–738.

Banerjee, A., 1999. Panel data unit roots and cointegration: an overview. Oxf. Bull.Econ. Stat. 61 (S1), 607–629.

Carrion-i-Silvestre, J.L., del Barrio-Castro, T., Lopez-Baso, E., 2005. Breaking thepanels: an application to GDP per capita. Econ. J. 8, 159–175.

Energy Information Administration, 2012. International Energy Outlook. EIA,Washington, DC, Available from: ⟨www.eia.gov/forecasts/aeo⟩.

Engle, R.F., Granger, C.W.J., 1987. Co-integration and error correction: representa-tion, estimation, and testing. Econometrica 55, 251–276.

Hadri, K., 2000. Testing for stationarity in heterogeneous panel data. Econ. J. 3,148–161.

Heston, A., Summers, R., Aten, B. 2012. Penn World Table Version 7.1. Center ofComparisons of Production, Income and Prices at the University of Pennsyl-vania. Accessed from: ⟨https://pwt.sas.upenn.edu/php_site/pwt71/pwt71_form.php⟩.

Kao, C., 1999. Spurious regression and residual-based tests for cointegration inpanel data. J. Econometr. 90, 1–44.

Lean, H.H., Smyth, R., 2010a. Multivariate Granger causality between electricitygeneration, exports and GDP in Malaysia. Energy 35, 3640–3648.

Lean, H.H., Smyth, R., 2010b. On the dynamics of aggregate output, electricityconsumption and exports in Malaysia: Evidence from multivariate Grangercausality tests. Appl. Energy 87, 1963–1971.

Levin, A., Lin, C.F., Chu, C., 2002. Unit root tests in panel data: asymptotic and finitesample properties. J. Econ. 108, 1–24.

Maddala, G.S., Wu, S.A., 1999. Comparative study of unit root tests with panel dataand a new simple test. Oxf. Bull. Econ. Stat. 108, 1–24.

Narayan, P.K., Smyth, R., 2009. Multivariate Granger causality between electricityconsumption, exports and GDP: evidence from a panel of Middle Easterncountries. Energy Policy 37, 229–236.

Pedroni, P., 2001. Purchasing power parity tests in cointegrated panels. Rev. Econ.Stat. 83, 727–731.

Pedroni, P., 2004. Panel cointegration: asymptotic and finite sample properties ofpooled time series tests with an application to the PPP Hypothesis. Econometr.Theor. 20, 597–625.

Sadorsky, P., 2009a. Renewable energy consumption, CO2 emissions and oil pricesin the G7 countries. Energy Econ. 31, 456–462.

Sadorsky, P., 2009b. Renewable energy consumption and income in emergingeconomies. Energy Policy 37, 4021–4028.

Sadorsky, P., 2011. Trade and energy consumption in the Middle East. Energy Econ.33, 739–749.

Sadorsky, P., 2012. Energy consumption, output and trade in South America. EnergyEcon. 34, 476–488.

United Nations Economic Commission for Africa (Office of North Africa), 2012. Therenewable Energy Sector in North Africa.

United Nations Industrial Development Organization, 2009. Scaling up RenewableEnergy in Africa.

World Bank, 2010. World Development Indicators. Accessed from: ⟨http://www.worldbank.org/data/onlinedatabases/onlinedatabases.html⟩.

World Trade Organization, 2011. Harnessing trade for sustainable development anda green economy.

Table 10Panel OLS, FMOLS and DOLS long-run estimates (model with imports).

Variables rec im k l

OLS 0.052444 (0.0001)nnn 0.208838 (0.0000)nnn 0.508536 (0.0000)nnn 0.175659 (0.0000)nnn

FMOLS 0.053928 (0.0500)nn 0.214332 (0.0012)nnn 0.515070 (0.0000)nnn 0.163482 (0.0008)nnn

DOLS 0.052444 (0.0608)n 0.208838 (0.0019)nnn 0.508536 (0.0000)nnn 0.175659 (0.0004)nnn

Cointegrating equation deterministics: intercept and trend.All the variables are measured in natural logarithms.P-values listed in parentheses.

nnn Critical values at the 1% significance level.nn Critical values at the 5% significance level.n Critical values at the 10% significance level.



http://refhub.elsevier.com/S0301-4215(13)01130-0/sbref1













www.eia.gov/forecasts/aeo





https://pwt.sas.upenn.edu/php_site/pwt71/pwt71_form.php

https://pwt.sas.upenn.edu/php_site/pwt71/pwt71_form.php





























http://www.worldbank.org/data/onlinedatabases/onlinedatabases.html

http://www.worldbank.org/data/onlinedatabases/onlinedatabases.html





output, renewable energy consumption and trade in africa

Documents