Renewable and Sustainable Energy Reviews 76 (2017) 620–627

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

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

Renewable energy consumption, International trade, oil price and economic MARK
growth inter-linkages: The case of Tunisia
⁎
Riadh Brinia, Mohamed Amarab, , Hatem Jemmalic
a
Faculty of Economic Sciences and Management of Nabeul, University of Carthage, and International Interaction Research Center (LIEI), University of
Tunis El-Manar, Tunisia
b
Higher School of Economic and Commercial Sciences of Tunis, University of Tunis, Tunisia
c
Higher Institute of Accounting and Business Administration, University of Manouba, and Laboratory for Research on Quantitative Development Economics
(LAREQUAD), University of Tunis El-Manar, B.P. 248 El Manar II, 2092 Tunis, Tunisia

A R T I C L E I N F O A BS T RAC T

Keywords: This paper contributes to the small but growing literature on the linkages between renewable energy
Renewable energy consumption consumption, international trade, oil price and economic growth. It aims to investigate such dynamic
International trade relationships using the bounds testing approach to cointegration and the ARDL methodology for Tunisia over
Oil price the period 1980–2011. The main empirical findings reveal the presence of a bidirectional relationship between
Economic growth
renewable energy consumption and international trade in the short-run. Indeed, an increase in oil price may
ARDL
Tunisia
imply an increase of renewable energy consumption. Furthermore, a unidirectional relationship between
renewable energy and oil price is proven in the short-run.

1. . Introduction responsible for the contamination of used data series. Using data from
six industrialized countries and with a time period of 30 years (1952–
Over the past four decades, the lion's share of the literature on 1982), Erol, Yu [13] found feedback causality running between
energy consumption has focused on the causal relationship between economic growth (EG) and energy consumption (EC) for Japan,
energy consumption and economic growth. Four testable hypotheses unidirectional causality from EC to EG for Canada, from EG to EC
can be distinguished to examine such relationship: growth, conserva- for Germany and Italy, and none for France and England.
tion, feedback, and neutrality hypothesis [2,30,41,38]. The growth Focusing on a set of developing countries, Masih and Masih [26]
hypothesis regards economy strongly linked to energy. According to postulated unidirectional causality from EC to EG in India, and from
this assumption, any reduction in energy use will lower automatically EG to EC in Pakistan and Indonesia, but none for Malaysia, Singapore
economic growth. Under the second hypothesis (conservation hypoth- and the Philippines. Soytas and Sari [40] postulated that economic
esis), it's assumed that unidirectional causality runs only from eco- growth Granger affects energy use in Italy and Korea, while a
nomic growth to energy use. Thus, any attempt to diminish energy unidirectional causality runs from energy use to economic growth in
consumption may not have much influence on economic growth. The other developed countries such as France, Germany, Japan and Turkey.
feedback hypothesis presumes the existence of bi-directional causality Huang et al. [16] supported the neutrality hypothesis for low-income
between energy consumption and economic growth. Under the fourth countries, but found unidirectional causality from EG to EC for middle
hypothesis (neutrality hypothesis), it's assumed that any change in and high-income countries, similarly to Aqeel and Butt [7], Shahbaz
energy consumption might not have any effect on economic growth, and Lean [39], Shahbaz and Feridun [37] for Pakistan, Lee [21] for
and vice versa [8]. France, Italy and Japan, and Lee and Chien [22] for France and Japan.
The first study on the growth-energy relationship was conducted by The reverse causal relationship running from EG to EC was postulated
Kraft and Kraft, [18] who found unidirectional causality from gross by Lee [21] for Canada, UK, Germany, Sweden, and Switzerland;
national product (GNP) growth to energy consumption using US data Narayan and Smyth (2008) for G-7 countries; Bowden and Payne [10]
from 1947 to 1974. Yet, reducing two years from the initial dataset, for the US.
Akarca and Long [1] did not find, surprisingly, any association between The disagreement among the aforementioned studies can be due
the two variables. They argue that the 1973 oil embargo is the mainly to methodological and data differences and countries hetero-

⁎
Correspondence to: 4, Rue Abou Zakaria El Hafsi 1089 Montfleury, Tunis, Tunisia.
E-mail addresses: riadh.brini@fsegs.rnu.tn (R. Brini), mohamed.amara.isg@gmail.com (M. Amara), hatemjemmali79@gmail.com (H. Jemmali).

http://dx.doi.org/10.1016/j.rser.2017.03.067
Received 26 January 2016; Received in revised form 20 January 2017; Accepted 10 March 2017
1364-0321/ © 2017 Elsevier Ltd. All rights reserved.
R. Brini et al. Renewable and Sustainable Energy Reviews 76 (2017) 620–627

geneity in climate, degree of development and energy use patterns. On Energy Conservation (ANME) today)1 that aims to develop a mechan-
the other hand, other studies have explored the relationship between ism for the rational use of energy as well as the promotion of renewable
EG and EC in the case of renewable energy (see for instance energies. This policy gained special momentum starting in the mid-
[3,4,5,6,38]). A consensus has emerged about the vital role of renew- 2000s, in the context of spiraling international oil prices and the
able energy consumption for increasing economic growth and reducing development of a steadily growing energy deficit [15]. The year 2000
greenhouse gas emissions [37]. was the turning point in the Tunisian's energy balance, where the
Recently, some centers and organizations such as the International country became a net importer of energy due to stagnation of the
Trade Centre (ITC), and the World Trade Organization (WTO) have production and the rapid growth of domestic demand (Fig. 1). As Fig. 1
shown an increasingly open approach towards economic growth, demonstrates, energy consumption increased at an average annual rate
energy and trade inter-linkages. The latest literature suggested that of 3.3% between 1995 and 2000, compared to an annual rate of 1.4%
trade openness can be involved in the production function to elucidate between 2005 and 2010. The demand recorded sustained growth,
the growth of gross domestic product (GDP). In this regard, several reaching 8.2 million toe in 2010, while the resources stagnated at
papers have emphasized on the relationship between total energy use, around 7.5 million toe.
trade and real GDP. Lean and Smyth [19] were the first to investigate Since 2009, primary energy consumption in Tunisia is dominated
the dynamic relationship between economic growth, electricity gen- by natural gas that reached 52.85 percent in 2014 and oil products with
eration, exports and prices for Malaysia. Their main findings from 46.62 percent share. Other primary energy sources, more specifically
Granger causality tests demonstrate the existence of a unidirectional the renewable energy, do not feature as significant energy inputs
causality running from economic growth to electricity generation. In (Fig. 2). Fig. 2 shows that, between 1990 and 2012, natural gas
their second study on Malaysia, Lean and Smyth [20] investigated the increased its share of energy consumption from 20 to 55 percent,
causal relationship between aggregate output, electricity consumption, while oil fell share decreased from 80 to 44 percent. The increase in the
exports, labor, and capital by using multivariate model. A feedback natural gas share can be explained by the fact that Tunisia receives
causality is found between aggregate output and electricity consump- natural gas from a pipeline between Algeria and Italy that runs across
tion. As policy implication, they recommend that Malaysia should its territory. In addition, significant natural gas deposits have recently
implement a dual strategy of rising investment in electricity infra- been discovered in southern Tunisia and in the Kerkenna Islands. The
structure and promoting electricity conservation policies to diminish hydrocarbons reserves at the end of the 2010 were estimated to be 430
the unnecessary wastage of electricity. million barrels of crude oil and 65 billion cubic meters of natural gas.
Similarly, Narayan and Smyth [28] found nearly the same conclu- The majority of Tunisia's natural gas output come from Miskar field
sion for a set of Middle East countries. A bidirectional effect was located about 125 km offshore in the Gulf of Gabes. According to
postulated between electricity consumption, exports and real GDP. British Gas (the largest investor in Tunisia's energy sector that owns
Using panel cointegration techniques, Sadorsky [35] demonstrated 100% interest in the Miskar field and 50% in the Hasdrubal field) the
how trade can influence energy consumption for eight Middle East Miskar field contains 1500 Bcf (about 42.5 ×108m3) of natural gas
countries. He found, in the short-run, a Granger causality running from reserves in Tunisia and provides more than 50% of Tunisia's total
exports to energy consumption, and a feedback relationship between natural gas demand [9].
imports and energy consumption. While in the long-run, he postulated In Tunisia, industrial sector is the largest energy end-user, which
that a raise in both exports and imports may influence the demand of accounted for 36% of total final energy consumption in 2010 followed
energy. In subsequent study, Sadorsky [36] strengthened the long-run by the transportation sector at 31%, buildings at 27% and finally
relationship between trade and energy consumption for a sample of agriculture at 6%. The distribution has not changed much compared to
seven South American countries. 1990: industrial sector 41%, transportation sector 30%; buildings 22%
The aim of this study is to investigate the relationship between and agriculture 7%.
renewable energy consumption, international trade, oil price and Aiming to reduce the economic burden of energy costs, the
economic growth for Tunisia for the period of 1980–2011 using government introduced a series of laws and programs to support the
autoregressive distributed lag (ARDL) approach of Pesaran et al., development of renewable energy and to increase the industrial energy
[31] and Granger causality tests. The current study extends the existing efficiency (RE & EE). Hence, Tunisia is bestowed with a high potential
empirical literature on the relationship and causality between renew- for both solar (more than 3,600 hours of sunshine per year) and wind
able energy consumption and economic growth for Tunisia. To our (7–10 m/s) energy due to its large coastal area and southern desert. In
knowledge, there have been no studies focusing on this causal relation- this framework, three main sustainable programs have been adopted
ship. In fact, Jebli and Ben Youssef [17] have investigated only the by the government: the 2005–2007 three-year program (accompanied
dynamic causal relationship between per capita CO2 emissions, real by the creation of National Fund for Energy Saving (FNME)), the
GDP, renewable and non renewable energy consumption and interna- 2008–2011 four-year program and the 2011–2016 Tunisian Solar
tional trade for the case of Tunisia, without considering the oil price Plan. The first program has resulted in the following actions [15]:
variable and the impact of aforementioned variables on economic About 230 EE program contracts were established between the
growth. In addition, to contributing to the broader empirical literature, government and companies in the industrial sector; Around
this paper has significant policy implications in the field of sustainable 120,000 m² of solar water heaters in the residential sector were
energy use. The remainder of the paper is structured as follows: Section installed (PROSOL program) and the distribution of around 1 million
2 gives an idea about the renewable energy sector, international trade energy-saving lamps in the residential sector. The impacts of this
and economic growth in Tunisia. Section 3 describes the data and the program are significant resulting to a cumulative energy saving of 770
methodology used. Section 4 deals with the empirical models and ktoe (where 92% from EE actions and 8% from renewable energy
results, and Section 5 concludes. development actions). Following those encouraging results, a new four-
year program for energy conservation for the period 2008–2011 was
2. Energy, international trade and economic growth in established with the main objective of increasing the share of renew-
Tunisia able in primary energy consumption of 4% in 2011 (ANME). In order
to strengthen national renewable energy policy, the government has
Since the mid-1980s, Tunisia has conducted a proactive policy for launched in 2009 the Tunisian Solar Plan (PST) covering the period
the promotion of renewable energy (RE) and energy efficiency (EE).
This was made through several initiatives including the creation in
1
1986 of the Agency of Energy Management (called National Agency for The ANME was established in 2004 under the Ministry of Industry.

621
R. Brini et al. Renewable and Sustainable Energy Reviews 76 (2017) 620–627

Fig. 1. Energy balance in Tunisia during 1990–2014, (data from the ANME).

Fig. 2. Final energy consumption by resource (1990–2014), (data from the ANME).

2011–2016. The plan aimed to increase the share of renewable energy supplying region and to control the innovative technologies in the
in electricity generation from 1 percent in 2012 to 11 percent in 2016 various fields of renewable energy. On the other side, the European
and around 40 percent in 2030, resulting in energy savings of 22% and Union views the energy cooperation with the Mediterranean Partners
a reduction of 1.3 million tones of CO2 per year. as an important tool in providing assistance with finance and technol-
In addition to meeting domestic demand through these programs, ogy transfer aimed at encouraging RE & EE in the Mediterranean
the strategic position of Tunisia on the power transmission lines regions. “This process will help rein in excessive growth in domestic
between Europe and North Africa raises the possibility that power demand and increase the energy available for export to Europe”
generated from RE projects could also be exported to Europe. To (Kaigiannas et al., 2003). This new and original Euro-Mediterranean
complement the efforts at national level and overcome all obstacles partnership between the southern-rim countries (with a significant
associated with weak institutional, technical and financial capacities, it potential for renewable energy) and the northern-rim countries (with
is important for Tunisia to strengthen intra-North Africa and North- most technology) is an opportunity to pursue the EE & RE objectives
South Mediterranean partnerships to promote the production and use more vigorously [11,14,25]. Therefore, there is a mutual interest
of renewable energy. North Africa, in general and Tunisia in particular, between the parties to develop the partnership. In this context, specific
has the comparative geographic and labor market advantages to extract Euro-Mediterranean platforms have been created such as the Euro-
an abundance of solar and wind energies. These potentials sources will Mediterranean Partnership (EMP) also known as the Barcelona
allow North African countries to maintain the position of an energy Process launched in 1995 and aimed to develop a multilateral approach

622
R. Brini et al. Renewable and Sustainable Energy Reviews 76 (2017) 620–627

Table 1 are order I(0) or order I(1). The ARDL model is also appropriate for
Descriptive statistics and correlation matrix. analyzing small sample size data and estimating the long- and short-
run relationship together. The ARDL approach leads to estimate the
Description LGDP LPO LERC LTRADE
following unrestricted error correction model by Ordinary Least
Mean 23.793 3.372 2.590 4.464 Squares Method (OLS):
Median 23.756 3.268 2.580 4.453
1 p q
Minimum 23.182 2.542 2.414 4.211 ΔLGDPt = α0GDP + ∑i =1 βiGDP ΔLGDPt − i + ∑ j1=0 χjREC ΔLRECt − j
Maximum 24.430 4.711 2.769 4.748
r
Std. Dev. 0.399 0.591 0.075 0.118 + ∑k1=0 δkTRADE ΔLTRADEt − k +
Skewness 0.114 0.765 0.644 0.091 s1
Kurtosis 1.721 2.566 2.538 3.180
∑l =0 ϕlPO ΔLPOt − l + φGDP LGDPt −1 + γREC LRECt −1 + ηTRADE LTRADEt −1
LGDP 1.000 + λPO LPOt −1 + ε1t
LPO 0.611 1.000
LERC −0.379 0.317 1.000 (2)
LTRADE 0.605 0.667 −0.049 1.000
2 q 2 p
Observations 32 32 32 32 ΔLRECt = α0REC + ∑ j =1 βjREC ΔLRECt − j + ∑i =0 χiGDP ΔLGDPt − i
r
+ ∑k2=0 δkTRADE ΔLTRADEt − k +
towards the region. In 2008 the Union for the Mediterranean (UfM) re- s2
∑l =0 ϕlPO ΔLPOt − l + φGDP LGDPt −1 + γREC LRECt −1 + ηTRADE LTRADEt −1
launched the idea of North-South energy cooperation, and particular
+ λPO LPOt −1 + ε2t
the Mediterranean Solar Plan (MSP) (Clombo and Sartori, 2014).
(3)
3. Data and methodology r p3
ΔLTRADEt = α0TRADE + ∑k3=1 βkTRADE ΔLTRADEt − k + ∑i =0 χiGDP ΔLGDPt − i
q
3.1. Data + ∑ j 3=0 δjREC ΔLRECt − j +
s3
∑l =0 ϕlpo ΔLPOt − l + φGDP LGDPt −1 + γREC LRECt −1 + ηCO2 LTRADEt −1
In this paper, we use annual data covering the period 1980–2011
which are collected from World Development Indicators (WDI-2014) + λPO LPOt −1 + ε3t
and statistical review of world energy (2015). The empirical analysis (4)
included real gross domestic product in constant US Dollars (GDP ),
4 s 4 p
combustible renewable and waste (% of total energy) in thousands of ΔLPOt = α0PO + ∑l =1 βsPO ΔPOt − l + ∑i =0 χiGDP ΔLGDPt − i
metric tons (ERC ), International trade defined as the sum of exports +
q
∑ j =0 δjREC ΔLRECt − j +
and imports divided by the GDP (TRADE ) and oil price in US dollars
r
per barrel (PO ). The ERC consumption includes the consumption of ∑k4=1 ϕkTRADE ΔLPOt − k + φGDP LGDPt −1 + γREC LRECt −1 + ηTRADE LTRADEt −1
hydropower, solar, wind, geothermal, bio-fuel and biomass. For the + λPO LPOt −1 + ε4t
stability of data, we use the natural logarithm for all variables. Table 1
(5)
shows some descriptive statistics of each series in the natural logarithm
form over the 1980–2011 period. The summary statistics include the where the Δ is the first difference and εjt , ( j = 1, 2, 3, 4) are white
means, median, maximum, minimum and the normality test. Fig. 3 noise error terms.
presents the evolution of each variable in Tunisia over the period The existence of cointegration relationship between variables from
1980–2011. As we can see the variation of the ERC decreased about Eq. 2 to Eq. 5 is examined by testing the significance of the lagged
20.5% between 1980–1992 while it has increased by 20% over the time levels of variables using the computed F-statistic. Pesaran et al., [31]
period 1992–2011. The lowest value of trade was 67% in 1986 (crisis propose testing (1) H 0 : φGDP = γREC = ηTRADE = λPO=0 which means
period in Tunisia) and it reached its highest value in 2008 (115%). The that we cannot reject the absence of cointegration, against the
oil price decreased by 65% between 1980–1998 and it has increased alternative (2) H1 : φGDP≠γREC ≠ηTRADE ≠λPO≠0 , which implies that the
from 13 USD per barrel in early 1998 to 97 USD in 2008, before falling hypothesis of the existence of such a relationship cannot be rejected.
to 62 USD per barrel by the end of that year (Subprime crisis). The oil Following Pesaran et al., [31], the F-statistic used for this test has a
price has increased again by about 79% over 2009–2011. The Tunisian non standard asymptotic distribution and generated two sets of critical
real GDP increased by 18% during the period 1980 to 1995 while it had values bounds (the lower critical value and the upper critical value).
increased by 70% over the period 1995–2011. The lower critical value corresponding to the case where all variables
are I(0) and upper critical value corresponding to the case where all
3.2. . Model specification variables are I(1). If the computed F-statistic exceeds the upper critical
bound, then the null hypothesis of the no cointegration is rejected and
Following Ocal and Aslan [29] and Lin and Moubarak [23], we use a we can conclude that there is evidence of a long-run relationship. If it
multivariate model in order to examine the relationship between ERC , falls below the lower critical value, we do not reject the null hypothesis
TRADE , PO and GDP . of no cointegration. Finally, if the F-statistic is between the lower and
upper critical bounds, the result is not conclusive.
LGDPt = f (LERCt , LPOt , LTRADEt ) (1)
Where LGDP is the natural log of real gross domestic product, LERC is 3.3. Granger causality test
the natural logarithm of renewable energy consumption, LPO is natural
log of the price oil and LTRADE is the natural log of sum of exports and In order to perform Granger-causality test, Pesaran et al., [32]
imports divided by the GDP . proposed to estimate a Vector Error Correction Model (VECM). Two
The Autoregressive Distributed Lag (ARDL) approach developed by steps of are used to investigate the long-run and short-run dynamic
Pesaran et al., [31] will be employed for cointegration. This approach relationships [12]. In the first step, the long-run parameters in Eq. 1
has many econometric advantages compared to the standard Johansen are estimated to obtain the residuals corresponding to the deviation
cointegration technique. The ARDL solves the endogeneity problem of from equilibrium. In the second step the parameters related to the
the explanatory variables, and it can avoid the problems of unit root short-run adjustment are estimated. The resulting equations employed
pre-testing as the test can be carried regardless of whether the series in conjunction with Granger causality testing are as follows:

623
R. Brini et al. Renewable and Sustainable Energy Reviews 76 (2017) 620–627

Fig. 3. Trends in renewable energy consumption, international trade, oil price and gross domestic product.

⎡ β1,1, i β1,2, i β1,3, i β1,4, i ⎤ Table 2

⎡ LGDP ⎤ ⎡ α1 ⎤ ⎢ ⎥ Unit root test.
⎢ ⎥ ⎢α ⎥
n
⎢β β β β2,4, i ⎥
(1 − L ) ⎢ LREC ⎥ = ⎢ α2 ⎥ + ∑ (1 − L ) ⎢ β2,1,i β2,2,i β2,3,i
⎢ LTRADE ⎥ 3 β3,4, i ⎥ variables ADF PP
⎣ LPO ⎦ ⎢⎣ α4 ⎥⎦ i =1 ⎢ 3,1, i 3,2, i 3,3, i ⎥
⎢⎣ β4,1, i β4,2, i β4,3, i β4,4, i ⎥⎦
Intercept Intercept and trend Intercept Intercept and trend
⎡ LGDPt − i ⎤ ⎡ γ1 ⎤ ⎡ ε1t ⎤
⎢ ⎥ ⎢γ ⎥ −2.87 −2.62
⎢ LRECt − i ⎥ + ⎢ 2 ⎥ [ECTt −1] + ⎢⎢ ε2t ⎥⎥
LGDP 0.566 0.19
∆LGDP −3.99*** −3.73** −6.34*** −6.32***
⎢ LTRADEt − i ⎥ ⎢ γ3 ⎥ ε3t
⎢⎣ ε4t ⎥⎦ LREC −1.88 −0.73 −2.05 −1.17
⎢⎣ LPOt − i ⎥⎦ ⎣ γ4 ⎦
∆LREC −3.34** −4.64*** −7.42*** −8.55***
(6) LTRADE −1.75 −3.24** −1.64 −2.57
∆LTRADE −4.08*** - −5.07*** −5.14***
Where α1, α2, α3, α4 constant trend, (1 − L ) is the difference operator, LPO 0.05 −1.38 −0.03 −1.31
γ1, γ2, γ3, γ4 are the adjustment coefficient and ε1, ε2 , ε3, ε4 are the ∆LPO −4.07*** −5.27*** −5.59*** −6.49***
Critical value
serially uncorrelated error terms. The ECTt−1 is the error correction
1% −3.67 −4.29 −3.66 −4.28
term which is a residual term obtained by running the long term 5% −2.96 −3.56 −2.96 −3.56
cointegration Eq. (1). The vector error correction model in Eq. (6) is 10% −2.62 −3.22 −2.61 −3.21
estimated with the same lag determined in unrestricted VAR structure.
The estimation of the long-run coefficient of the ARDL model using Notes: ∆ denotes the first difference operator. *** p-value < 0.01; ** p-value < 0.05; * p-
value < 0.1.
Eqs. (2)–(5) is based on three steps: In the first step, the order of
integration of all variables is determined by using the Augmented
Dickey-Fuller (ADF) and Phillips-Perron (PP) unit root tests. In the are nonstationary in their level, but stationary in their first difference at
second step, the long-run equilibrium relationship among the different 5% level, except the international trade variable that is stationary in its
variables is estimated. Finally, in the third step, the granger causality level form according the ADF test.
test is employed to estimate the causal relationship between different
variables. 4.2. ARDL cointegration tests

The first stage of the ARDL bound test is to select the optimal lag
4. Empirical results that minimize Final Prediction Error criterion (FPE), Akaike
Information Criterion (AIC), Schwarz Information Criterion (SIC)
4.1. Unit root test and Hannan Quinn criterion (HQ) of the unrestricted VAR model.
The result indicates that the optimal lag length is one (Table 3).
Table 2 presents the results of both ADF and PP tests. All variables Following this, we test the existence of cointegration relationship

624
R. Brini et al. Renewable and Sustainable Energy Reviews 76 (2017) 620–627

Table 3 Table 6
Lag length criteria. Order of optimal lags.

Lag LogL LR FPE AIC SIC HQ Number of lags Eq. (2) AIC SIC

0 72.027 NA 1.08e-07 −4.692 −4.503 −4.632 p1 = 2, q1 = 0, r1 = 0, s1 = 0 −5.219 −4.939

1 182.341 182.588a 1.64e-10a −11.196a −10.253a −10.901a p1 = 2, q1 = 0, r1 = 2, s1 = 0 −5.174 −4.800
2 193.360 15.198 2.48e-10 −10.852 −9.155 −10.321 p1 = 2, q1 = 0, r1 = 2, s1 = 1 −5.173 −4.752
3 211.726 20.266 2.54e-10 −11.016 −8.564 −10.248 p1 = 2, q1 = 0, r1 = 0, s1 = 1 −5.170 −4.843
p1 = 2, q1 = 0, r1 = 1, s1 = 0 −5.169 −4.842
a
Indicates lag order selected by the criterion. p1 = 2, q1 = 1, r1 = 0, s1 = 0 −5.166 −4.839
p1 = 1, q1 = 1, r1 = 1, s1 = 1 −5.156 −4.782
Table 4 p1 = 1, q1 = 1, r1 = 1, s1 = 2 −5.143 −4.723
Bounds cointegration test - ARDL approach. p1 = 2, q1 = 0, r1 = 1, s1 = 1 −5.141 −4.767
p1 = 2, q1 = 1, r1 = 1, s1 = 1 −5.135 −4.715
Equations used (2)–(5) F-statistics p1 = 1, q1 = 1, r1 = 2, s1 = 1 −5.134 −4.713
p1 = 2, q1 = 1, r1 = 2, s1 = 1 −5.131 −4.664
FLGDP (LGDP /LREC ,LTRADE , LPO ) 6.484***
p1 = 2, q1 = 1, r1 = 1, s1 = 0 −5.127 −4.754
FLREC (LREC /LGDP,LTRADE , LPO ) 2.535
p1 = 2, q1 = 1, r1 = 0, s1 = 1 −5.125 −4.751
FLPO (LPO /LGDP,LTRADE , LREC ) 2.374
2.374 p1 = 2, q1 = 1, r1 = 0, s1 = 2 −5.121 −4.748
FLTRADE (LTRADE /LREC ,LGDP, LPO )
Critical values of F-statistic I(0)=3.23 I(1)=4.35 p1 = 2, q1 = 1, r1 = 0, s1 = 0 −5.112 −4.692
p1 = 1, q1 = 2, r1 = 1, s1 = 1 −5.108 −4.688
Critical values for K=3 and n=30 at 5%. p1 = 2, q1 = 0, r1 = 2, s1 = 2 −5.106 −4.639
***
p-value < 0.01 p1 = 1, q1 = 1, r1 = 0, s1 = 1 −5.103 −4.776
p1 = 2, q1 = 2, r1 = 0, s1 = 0 −5.100 −4.727

Table 5
Johansen cointegration test.
Table 7
Hypothesized No. of Eigenvalue Trace 0.05 Critical Prob** ECM-ARDL results.
CE(s) statistic value
A. Error correction representation of B. long run coefficients:
None* 0.617 60.295 47.856 0.002 the ARDL (2.0.0.0). The dependent dependent variable is ΔLGDP
At most 1* 0.550 32.475 29.797 0.024 variable is ΔLGDP
At most 2 0.236 9.339 15.495 0.335
At most 3 0.052 1.541 3.841 0.215 Regressors Coefficients Prob. Regressors Coefficients Prob.

ΔLGDPt−1 −0.346 **
0.020 LREC −11.999 **
0.017
Hypothesized No. of Eigenvalue Trace 0.05 Critical Prob** ΔLREC −0.397*** 0.000 LTRADE −2.301 0.185
CE(s) statistic value ΔLTRADE −0.076* 0.051 LPO 1.175*** 0.008
ΔLPO 0.039*** 0.001 C 62.929*** 0.002
None* 0.617 60.295 47.856 0.002
ECTt−1 −0.033** 0.0385
At most 1* 0.550 32.475 29.797 0.024
LM test (p- 0.494
At most 2 0.236 9.339 15.495 0.335
value)
At most 3 0.052 1.541 3.841 0.215
ARCH test (p- 0.793
* value)
Denotes rejection of the hypothesis at the 0.05 level.
** RESET test 0.247
MacKinnon-Haug-Michelis [24] p-value.
(p-value)
Normality 0.639
among the variables using the bounds F-statistic developed by Narayan test (p-
[27]. The results of this test are reported in Table 4. When the gross value)
domestic product (GDP) is a dependent variable, the computed F- ***
p-value < 0.01.
statistic (6.484) is higher than the upper bound critical value (4.35) at **
p-value < 0.05.
5% significance level for 30 observations. This result suggests that there *
p-value < 0.1
is a cointegration relationship among economic growth, renewable
energy consumption, international trade and oil price. Nevertheless, that the coefficients associated with renewable energy consumption in
when the renewable energy consumption is a dependent variable, there natural logarithm have a long-run and short-run negative and sig-
is no cointegration relationship among the variables (the computed F- nificant impact on real GDP in natural logarithm. More specifically, a
statistic (2.535) is less than the lower bound critical value (4.35) at 5% 1% increase in the renewable energy consumption implies a decrease in
significance level). Furthermore, when the oil price and the interna- the real GDP by 11.99% over the long-run and 0.397% over the short-
tional trade are dependant variables, the F-statistic are less than the run. This finding is similar to the result provided by Ocal and Aslan
lower bound critical value (3.23), implying the absence of cointegration [29] for Turkey. Ocal and Aslan [29] argued that in a developing
relationship between the variables. In order to reinforce the result country, economic growth can be adversely affected by renewable
obtained from bounds F-statistic, we apply Johansen cointegration energy consumption due to mismanagement of the energy sources,
test. The results presented in Table 5 confirm the existence of one long such as failing to use the rich sources. In fact, despite their availability
run relationship among the variables. and continuity, the renewable energy sources require significant
After identifying the co-integration relationship, we precede now to investment to be as energy sources, which means that they are
estimate the long-run relationship between variables using the ARDL expensive investments.
model. The order of optimal lags ( p1, q1, r1, s1) on the first differenced The estimated coefficients of regression found that in the long-run and
variables for Eq. (2) is obtained by minimizing AIC and SIC criterions. in the sort-run oil price has a positive and statically significant effect on
For the Eq. 2 the minimization of the two criterions leads to choose the real GDP . It means that an increase in the oil price by 1% leads an
optimal lags: p1 = 2, q1 = 0, r1 = 0, s1 = 0 (Table 6). increase in economic growth by 1.175% over the long-run and by 0.039%
Eq. (2) is estimated to obtain the long-run and short-run coeffi- over the short-run. Both of which are significant at the 1% level. This
cients using ARDL approach (Table 7). The results in Table 7 indicate

625
R. Brini et al. Renewable and Sustainable Energy Reviews 76 (2017) 620–627

Table 8
Granger causality test: Wald test.

F-Statistics

∆LGDPt−1 ∆LRECt−1 ∆LPOt−1 ∆LTRADEt−1 ECTt−1 FigureObject

∆LGDP – 0.195 (0.177) -0.015 (0.480) -0.027 (0.688) -0.024*** (0.002)

∆LREC 0.991** (0.014) – -0.054 (0.144) 0.213** (0.046) –
∆LPO -2.190 (0.395) 0.847 (0.540) – 0.590 (0.359) –
∆LTRADE 0.558 (0.535) 0.512 (0.227) -0.021 (0.758) – –

Notes: p-value in parenthesis. *** p-value < 0.01; ** p-value < 0.05; * p-value < 0.1.

Fig. 4. Plots of CUCUM and CUSUM Square.

result might be explained by the existence of oil demand shocks. Uhlig 4.3. Granger causality
[42] identifies the oil shocks as oil demand and oil supply shocks and
argues that the oil demand shocks affect positively the real GDP . The Table 8 presents the result of the Granger causality analysis using
results, also, show a negative and significant coefficient of international the VECM. In the short-run there is no evidence of granger causality
trade in the short-run but insignificant effect in the long-run. This running renewable energy consumption, oil price and international
indicates that in the long-run, the international trade did not lead to an trade to output. However, there is a short-run causality running output
increase output. Similarly, Balloumi (2014) reveals that the relation and international trade to renewable energy consumption. This result
between international trade and real GDP is insignificant in the long-term. means that, in the short-run, variation in output and international
The lag error correction term (ECTt−1) represents the speed of trade will impact renewable energy consumption. In addition, there is
adjustment of real GDP to its long-run equilibrium following a shock. no evidence of short-run causality running renewable energy consump-
The coefficient of −0.033 is negative and significant at the 5% level. tion, oil price and output to international trade.
This results indicate the existence of a stable long-run relationship The lag error term (ECTt−1) of the renewable energy consumption,
between LREC , LPO, LTRADE and LGDP . The same results suggest international trade and oil price are negative and statistically signifi-
that a deviation from the long-run equilibrium level of real GDP in 1 cant. This confirms that in the long-run there is Granger causality from
year is corrected by 3.3% in the next year. renewable energy consumption, international trade, oil price to eco-
The diagnostic tests of the ARDL model prove the absence of serial nomic growth.
correlation (LM test), the absence of heteroscedasticity (ARCH test), To summarize, in the long-run there is a unidirectional causal
the functional errors is specified (RESET test) and the normality of the relationship among renewable energy consumption to economic
residuals. growth, and there is unidirectional short run causality running from

626
R. Brini et al. Renewable and Sustainable Energy Reviews 76 (2017) 620–627

economic growth to renewable energy consumption. This confirms the nexus: evidence from a panel error correction model. Energy Econ 2012;34:733–8.
[7] Aqeel A, Butt MS. The relationship between energy consumption and economic
conservation hypothesis. Also, there is a short-run unidirectional growth in Pakistan. Asia-Pacific Dev J 2001;8:101–10.
causality running from international trade to renewable energy con- [8] Belke A, Dobnik F, Dreger C. Energy consumption and economic growth: new
sumption. This means that variation in international trade will affect insights into the cointegration relationship. Energy Econ 2011;33(5):782–9.
[9] Belloumi M. Energy consumption and GDP in Tunisia: cointegration and causality
renewable energy consumption. However, in the short-run there is no analysis. Energy policy 2009;37(7):2745–53.
evidence of Granger causality running from renewable energy con- [10] Bowden N, Payne JE. The causal relationship between US energy consumption and
sumption to oil price. In this case, the hypothesis of neutrality is real output: a disaggregated analysis. J Policy Modeling 2009;31:180–8.
[11] Darbouche H. Third time lucky* Euro-Mediterranean energy co-operation under
supported. the union for the Mediterranean. Mediterranean Politics 2011;16(01):193–211.
In order to check the robustness of the results, we apply the [12] Engle RF, Granger CW. Co-integration and error correction: representation,
cumulative sum (CUSUM) and cumulative sum squares (CUSUMS) of estimation, and testing. Econ: J Econ So 1987:251–76.
[13] Erol U, Yu ESH. On the causal relationship between energy and income for
recursive residuals tests. These tests show that there is no serial
industrialized countries. J Energy Dev 1987;13:113–22.
correlation and no problem of heteroscedasticity (Fig. 4). Fig. 4 proves [14] Escribano G. Convergence towards differentiation: the case of Mediterranean
that the curve is within the critical interval at 5% significance. This energy corridors. Mediterranean Politics 2010;15(2):211–29.
result indicates that coefficients are stable over the study period and [15] GIZ. Renewable Energy and Energy Efficiency in Tunisia–Employment
Qualification and Economic Effects, 2012. Retrieved from: 〈http://www.giz.de/
the model is valid. expertise/downloads/giz2012-en-employment-renewable-energy-tunisia.pdf〉.
[16] Huang B, Hwang MJ, Yang CW. Causal relationship between energy consumption
5. Conclusion and policy recommendations and GDP growth revisited: a dynamic panel data approach. Ecol Econ
2008;67:41–54.
[17] Jebli MB, Ben Youssef S. The environmental Kuznets curve, economic growth,
Renewable energy source has acquired great importance in today′s renewable and non-renewable energy, and trade in Tunisia. Renew Sustain Energy
world due to its cleanness and less negative environmental impact. Rev 2015;47:173–85.
[18] Kraft J, Kraft A. On the relationship between energy and GNP. J Energy Dev
Besides, countries using these clean sources will be less associated on 1978;3:401–3.
imported fossil fuels and accordingly less impacted by economic crisis. [19] Lean HH, Smyth R. CO2 emissions, electricity consumption and output in ASEAN.
For this reason, all developing and developed countries must increase Appl Energy 2010;87:1858–64.
[20] Lean HH, Smyth R. Multivariate Granger causality between electricity generation,
their activities for both consumption and production of renewable
exports and GDP in Malaysia. Energy 2010;35:3640–8.
energy. [21] Lee C. The causality relationship between energy consumption and GDP in G-11
Tunisia is considered among the world's poor in fossil fuel produc- countries revisited. Energy Policy 2006;34:1086–93.
[22] Lee C, Chien M. Dynamic modelling of energy consumption, capital stock, and real
tion, but it is leader in sources of renewable energy (solar, wind,
income in G-7 countries. Energy Econ 2010;32:564–81.
geothermal…). These potentials sources will allow Tunisia to maintain [23] Lin B, Moubarak M. Renewable energy consumption–economic growth nexus for
the position of an energy supplying country and to control the China. Renew Sustain Energy Review 2014;40:111–7.
innovative technologies in the various fields of renewable energy. [24] MacKinnon JG, Haug AA, Michelis L. Numerical distribution functions of like-
lihood ratio tests for cointegration. J Appl Econ 1999;14:563–77.
This paper examines the long-run and the short-run dynamic relation- [25] Marks J. High hopes and low motives: the new euro‐mediterranean partnership
ship and causality between renewable energy consumption and eco- initiative. Mediterranean Politics 1996;1(1):1–24.
nomic growth in Tunisia by including international trade and oil price [26] Masih AMM, Masih R. Energy consumption, real income and temporal causality:
results from a multi-country study based on cointegration and error-correction
over the period 1980–2011. Although, the literature has recently been modeling techniques. Energy Econ 1996;18:165–83.
conducted on the relationship and causality between economic growth [27] Narayan PK. The saving and investment nexus for China: evidence from coin-
and renewable energy, there is no study which has investigated this tegration tests. Appl Econ 2005;37(17):1979–90.
[28] Narayan PK, Smyth R. Multivariate Granger causality between electricity con-
relation in Tunisia. sumption, exports and GDP: evidence from a panel of Middle Eastern countries.
Using the ARDL approach and Granger causality test, our empirical Energy Policy 2009;37:229–36.
results show the negative impact of renewable energy consumption on [29] Ocal O, Aslan A. Renewable energy consumption–economic growth nexus in
Turkey. Renew Sustain Energy Rev 2013;28:494–9.
the economic growth and the unidirectional long-run causality running [30] Ozturk I. A literature survey on energy-growth nexus. Energy Policy
from renewable energy consumption to economic growth. This study 2010;38:340–9.
suggests the conservation hypothesis for the relationship between [31] Pesaran MH, Shin Y, Smith RJ. Bounds testing approaches to the analysis of level
relationships. J Appl Econ 2001;16(3):289–326.
renewable energy consumption and economic growth in Tunisia.
[32] Pesaran MH, Shin Y, Smith RP. Pooled mean group estimation of dynamic
More specifically, in the short-run there is a unidirectional relationship heterogeneous panels. J Am Stat Assoc 1999;94(446):621–34.
between international trade, economic growth and renewable energy [33] Sadorsky P. Renewable energy consumption CO2 emissions and oil prices in the G7
consumption. Trade enables Tunisia to benefit from technology countries. Energy Econ. 2009;31:456–62.
[34] Sadorsky P. Renewable energy consumption and income in emerging economies.
transfer that helps to invest more in the renewable energy sector. Energy Policy 2009;37:4021–8.
However, our results show the neutrality hypothesis in the short run [35] Sadorsky P. Trade and energy consumption in the Middle East. Energy Econ
between renewable energy consumption and oil price. 2011;33:739–49.
[36] Sadorsky P. Energy consumption, output and trade in South America. Energy Econ
2012;34:476–88.
References [37] Shahbaz M, Feridun M. Electricity consumption and economic growth empirical
evidence from Pakistan. Quality Quantity 2012;46:1583–99.
[38] Shahbaz M, Khan S, Tahir MI. The dynamic links between energy consumption,
[1] Akarca AT, Long TV. On the relationship between energy and GNP: a reexamina-
economic growth, financial development and trade in China: fresh evidence from
tion. J Energy Dev 1980;5:326–31.
multivariate framework analysis. Energy Econ 2013;40:8–21.
[2] Apergis N, Payne JE. Energy consumption and economic growth: evidence from the
[39] Shahbaz M, Lean HH. Does financial development increase energy consumption*
commonwealth of independent states. Energy Econ 2009;31:641–7.
The role of industrialization and urbanization in Tunisia. Energy Policy
[3] Apergis N, Payne JE. The emissions, energy consumption, and growth nexus:
2012;40:473–9.
evidence from the common wealth of independent states. Energy Policy
[40] Soytas U, Sari R. Energy consumption and GDP: causality relationship in G-7
2010;38(1):650–5.
countries and emerging markets. Energy Econ 2003;25:33–7.
[4] Apergis N, Payne JE. Renewable energy consumption and economic growth:
[41] Squalli J. Electricity consumption and economic growth: bounds and causality
evidence from a panel of OECD countries. Energy Policy 2010;38:656–60.
analyses for OPEC members. Energy Econ 2007;29:1192–205.
[5] Apergis N, Payne JE. The renewable energy consumption-growth nexus in Central
[42] Uhlig H. What are the effects of monetary policy on output* Results from an
America. Appl Energy 2011;88:343–7.
agnostic identification procedure. J Monetary Econ 2005;52:381–419.
[6] Apergis N, Payne JE. Renewable and non-renewable energy consumption-growth

627