By Nadia Ouédraogo, PhD Student University of Paris-Dauphine 1

ENERGY AND ECONOMIC POVERTY: AN ASSESSMENT BY STUDYING THE CAUSALITY BETWEEN ENERGY CONSUMPTION AND ECONOMIC GROWTH IN THE ECONOMIC COMMUNITY OF WEST AFRICA STATES (ECOWAS) By Nadia Ouédraogo, PhD Student University of Paris-Dauphine 1 Abstract The increasing attention given to global energy issues and the international policies needed to reduce greenhouse gas emissions have given a renewed stimulus to research interest in the linkages between the energy sector and economic performance at country level. Economic poverty is linked with energy poverty and, at the same time, energy is an important vector for triggering economic development and for reaching the objectives of the Millennium Development Goals (MDGs). In this paper we attempted to find the direction of the causal relationship between energy consumption and economic activity in the ECOWAS. More specifically we investigated the causal relationship between growth in energy consumption and growth in GDP. Additionally, to explore the possibility of further information on the direction of causality we disaggregated energy consumption into its components of petroleum and electricity consumption. The methodology was based on the Granger causality test which has been found appropriate by using the cointegration technique. 1 Contact details: Ouedraogo S. Nadia Center for Geopolitics of Energy and Raw Materials (CGEMP), Paris Dauphine University, Place du Maréchal de Lattre de Tassigny 75116 Paris Cedex 16 Tel: + 33 6 81 04 52 03 Email: souedraogo@dauphine.fr, nadiouchckaaa@yahoo.fr

I. Introduction Energy poverty is always associated with economic poverty. It concerns people that have a low income, a low energy consumption and no access, or limited access to modern energy fuel (petroleum products and electricity). Around 1.6 billion people do not have access to modern energy fuels (the majority of whom lives in Sub Saharan). It means that they don t have access to economic development and they spend a good deal of their time collecting local energy resources such as wood and dung, causing health disease and accelerating deforestation. This situation is worse by the recent episode of high oil price (the price of the oil barrel went up from $75 in 2002 to $146 in June 2008) caused some turmoil in the weak economies mainly in the ECOWAS: a region that suffers deeply from the changing price of oil. Energy poverty can be defined as "the absence of sufficient choice that allows access to adequate energy services, affordable, reliable, effective and sustainable in environmental terms to support the economic and human development"(reddy, 2000). From this definition, energy poverty is an obstacle to economic development but energy poverty is basically explained by a low income situation. The purpose of this paper is to understand more clearly the link between consumption and economic development and, and then, to see how energy poverty reduction may trigger economic development. To do this, we analyse the causal relationship between economy and energy by adopting a Vector Error Correction Model for non-stationary and cointegrated data with a sample of the fifteen countries of the ECOWAS and three sub-components of energy (total primary energy-include biomass, oil products and electricity). What is the impact of these sub-components of energy in economic development of the ECOWAS? This impact is it different depending on the type of energy? What impact of high oil price on energy consumption and economic development? What type of policy can be set up to overcome energy poverty and trigger economic development in the ECOWAS? The paper is divided into three sections. The first covers energy poverty and its impact on economic poverty of the region. The second section, discusses the link between energy consumption and economic growth. Based on these results, the third section seeks to determine the required energy policy and strategy to reduce the energy vulnerability and to fight against poverty and for sustainable development of the ECOWAS. II. Economic poverty and energy poverty At the beginning of the 3rd millennium, the energy situation in sub-saharan Africa, which closely reflects the low level of development, can be characterized as "energy poverty". Energy poverty can be defined as "the absence of sufficient choice that allows access to adequate energy services, affordable, reliable, effective and sustainable in environmental terms to support the economic and human development"(reddy, 2000). From

this definition, energy poverty is an obstacle to economic development but energy poverty is basically explained by a low income situation. World energy consumption is very unevenly. Per capita energy consumption is 0.5 ton of oil equivalent (toe) in SSA, 1 toe in China, 4 in Europe and 8 in the United States. Per capita consumption of commercial energy in the United States is 80 times higher than in Africa, 40 times higher than in South Asia, 15 times higher than in East Asia, and 8 times higher than in Latin America. Parts of world are over-energetized, others are under-energetized. For a population estimated in 2008 to 808 million people representing 11% of world population, Africa has consumed 469 million tonnes of oil equivalent (or 5.7% of available energy on the globe) energy primary in 2008. A quantity which gives a per capita the lowest in the world, with 0.59 toe per year against 1.76 toe in global average (4.31 toe for Western Europe, 8.46 toe for North America). According the Business As Usual scenarios, the situation will get worse in 2030: the average energy consumption of Africa should continue to fall (from 0.47 toe per year) due to population growth particularly strong on the mainland (+1.9 % per year, cons + 1% worldwide). Its total consumption should she go to 687 million toe spread between 1.5 billion people (or 5.9% of global consumption to 17.6% of the population of the planet). Less than 10% of the SSA rural population has access to modern energy services. Biomass provides over 80% of total domestic primary energy supply across the region even in major petroleum exporting countries. The remaining 20% of the energy requirement is mainly met by imported petroleum products. The high dependence of SSA on oil imports makes it extreme vulnerable to energy price shocks. Thereby, the recent episode of high oil prices is having devastating effects on the region and may undermine its good progress towards the economic development. The Economic Community of West African States ECOWAS groups 15 countries since its creation in 1975. Its size is 5,112 km 2 and its total population today stands at 262 million, representing 40% of the total population of sub-saharan Africa (SSA). With 7 % of the world GDP (161 billion current $ US) and $615 GDP per capita, 13 of 15 ECOWAS nations are currently categorized as Least Developed Countries (LDCs). Only two countries, the Cape Verde Islands and Ghana, are classified as countries with a medium level of development. All these nations also belong to the Heavily Indebted Poor Countries (HIPCs), and 14 of them have low levels of human development indicator (HDI lower than 0.5) and high poverty rates. West Africa has a population of 100 million poor, which means 44% of the region s population lives below the monetary poverty line of $1 per person per day. This is the highest percentage of any region in the world. Worse still, the figure is rising steadily and all signs indicate that the poverty reduction goal will not be met in this region by 2015 2. 2 World Bank (2007).

The energy consumption levels of people in the ECOWAS reflect the situation of energy poverty specifically relates to rural and suburban areas. With 4 % of the world population, (13% for Africa) and a 2 % production of world commercial energy (7% for Africa), ECOWAS accounts for only 1.7 % of world energy consumption (3% for Africa) 3. Biomass accounts for more than 80% of total energy consumption for domestic purposes (Enerdata, 2005). The intensive use of biomass has many negative effects on economic development, health and the environment. Loss of time: women and children spend several hours a day collecting firewood and other biomass. This is often detrimental to productive activities and education. The health problems: the use of biomass furnaces and traditional homes is not only inefficient (very low yields) but also constitutes a danger to the health of users. Wood combustion is incomplete and releases of toxic gas which is exacerbated by the lack of ventilation due to high indoor air pollution (Bruce et al., 2000). This pollution causes respiratory diseases, cancer, tuberculosis, and causes weight loss and eye diseases among newborns (Bruce et al., 2000). Women and children are the first concerned by this problem. Environment: the collection and use of fuelwood is another cause of deforestation, erosion, land degradation and desertification. As forests are being cleared for agriculture and other purposes, women and children are walking longer distances to collect biomass. Furthermore, the use of dung for heat production reduces the amount available for fertilizer production. Another illustration of the energy poverty of the continent is provided by this modern form of energy that is the ultimate power as shown by the following indicators: The power consumption per capita in ECOWAS is 128 kwh per capita. This is about 100 times less than in the USA (12.331 kwh per capita). Four out of the community s 15 countries have an overall electricity access rate ranging from 20 to 40%; these are Benin, Senegal, Ghana and Cote d Ivoire. There are considerable Service access gaps among urban zones (average 40%) and rural areas (average 6% to 8%) Despite the active promotion of privatization and competition reforms by international financial agencies such as the World Bank, the electricity sectors of ECOWAS countries remain highly vertically integrated and under state-ownership. Only two ECOWAS countries (Cape Verde and Cote d Ivoire) have a power sector owned in majority by (foreign) private companies (Pineau, 2007). In terms of national sector integration, all countries remain vertically integrated, except for three countries with independent distribution companies: Benin, Ghana and Togo. Only five out of the 15 ECOWAS countries have independent power producers (IPPs): Burkina Faso, Cote d Ivoire, Ghana, Nigeria and Senegal (Besant-Jones, 2006). 3 AIE, 2004 and Enerdata, 2005

The power sector is therefore still characterized by: excessive costs, low service quality, poor investment decisions, and lack of innovation in supplying customers (Besant-Jones). This contributes to the extremely low electricity usage in these regions. The low consumption levels show that there is insufficient access to energy services which does not allow for the development of economic activities or access to basic social services and as such, does not contribute to poverty reduction. There figures are however important differences among countries within ECOWAS. These differences reflect the important local conditions among countries and the difficulty to make general conclusions about energy consumption and income levels. III. Energy and economic development 1. Literature review The causal relationship between energy consumption and income is a well-studied topic in the literature of energy economics. The causality is in the sense of Granger causality (Granger, 1969). Kraft and Kraft (1978), in their pioneering study, found unidirectional causality running from GNP to energy consumption for the United States. They utilized the technique of Sims (1972) and used annual data for the 1947-1974 period. However, Akarca and Long (1980) pointed out that the Kraft and Kraft results are spurious by changing the time period by 2 years. Other studies utilizing different time periods and different techniques have either confirmed or contradicted Kraft-Kraft results (Abosedra and Baghestani, 1989; Yu and Choi, 1985; Cheng, 1995; Hwang and Gum, 1991; Erol and Yu, 1987). The large number of studies in this area, unfortunately, found different results for different countries as well as for different time periods within the same country. In most recent studies, researchers have focused on the cointegrating relationship between energy consumption and income for a few countries (e.g. Yu and Jin, 1992; Masih and Masih, 1996, 1997; Glasure and Lee, 1997).

Table 1: The comparison of empirical results from causality tests for developing countries Notes: Energy Income denotes causality runs from energy consumption to income. Income Energy denotes causality runs from income to energy consumption. Energy Income denotes bi-directional causality between income and energy consumption. Source: C.-C. Lee / Energy Economics 27 (2005) 415 427 In many developing countries the estimation of causality cannot be achieved, because of a short data span, which lowers the power of the unit root and cointegration tests. In this paper we are trying to reexamine the causal relationship between GDP and energy consumption in the ECOWAS countries, using cointegration and vector error correction techniques. But because of the lack of data, the results are not always strong. We use annual energy, electricity and oil consumption (ener, elec and oil hereafter) and GDP. Energy and oil use are a million metric tons of oil equivalents, electricity in million watt-hours and are sourced from Enerdata. The GDP variables are in millions and sourced in World Development Indicators (World Bank). For all countries, the time period used is 1980-2008 and all variables are in log.

2. Methodology Granger-causality implies causality in the prediction (forecast) sense rather than in a structural sense. It starts with the premise that the future cannot cause the past ; if event A occurs after event B, then A cannot cause B (Granger 1969). This concept can be examined in the context of a bivariate model consisting of the following two equations: (1) ET (2) where: e t =ln(e t ) y t =ln(y t ) Et = energy consumption; and Yt = GDP Following the development of unit root testing and cointegration, for non-stationary variables, integrated of order one or I(1), this initial formulation by Granger used levels of variables as shown in equations (1) and (2)are replaced by: (3) and (4) where Δ is the first difference operator, so that the terms are introduced in differences to ensure that they are stationary or I(0). Here the concept of causality is formulated in terms of changes to the variables and the presence of Granger-causality depends on the significance of the Δe t-j terms and Δy t-j terms in equations (3) and (4) respectively.

Furthermore, if it is found that the two integrated variables co-integrate, then equations (3) and (4) can be augmented as follows: (5) And (6) Where EC is the error correction term from a cointegrating equation of the form yt = βet +ECt and hence is I(0) 4. In this formulation there are two possible sources of Granger-causality, either through the lagged Δe terms if δj 0 or through the ECt 1 term, if σ2 0 (implying a long run relationship). Stage 1: Test the stationarity of the variables for each country using the Phillips-Perron test. If both e and y are I(1) then proceed to Stage 2. If one or both are not I(1) proceed to Stage 3a. Stage 2: Test for cointegration between e and y using the Johansen technique. If cointegration is not found proceed to Stage 3a. If cointegration is found proceed to Stage 3b. Stage 3a: Test for causality from e to y using the Granger causality procedure. Stage 3b: A long run relationship exists so there must be causality for at least one direction. Therefore, test if it is from e to y (or from y to e) using the error correction Equation and test accordingly. 3. Empirical results and discussion 3 1. Unit Root test We use the Phillips Perron (1988) method to test for the existence of unit roots and identify the order of integration for each variable. Unit root tests are done both with and without allowing for a time trend. The Newey and West (1987) method is applied to choose optimal lag lengths, which turned out to be 3 for all variables. As shown in Table 2, a unit root can be rejected for the first difference but not the levels for all variables at the 5% significance level. 4 In essence, if a pair of I(1) series are co-integrated, there must be Granger-causality in at least one direction (either e y and/or y e) hence it is necessary to add the EC term in Equation 5&6 to avoid miss-specifying the model and missing one source of causality.

Using these results, in the next step, we utilized Johansen (1988) and Johansen and Juselius (1990) maximum likelihood procedure to test for cointegration. 3. 2 Cointegration Test for cointegration between e and y using the Johansen technique. For consistency, the specification that allows for a linear trend in the data with an intercept but no trend in the co-integrating vector is utilised with the optimal lag structure for the VAR selected by using the Schwarz (SIC) criteria 5. Cointegration is accepted if the Trace and Max-eigenvalue test statistics indicate a cointegrating vector at the 5% level of significance. The results provide evidence of a cointegrating vector for 8 of the 15 countries (Table 3). For six of these eight countries (Benin, Guinea Bissau, Liberia, Mali Nigeria, Sierra Leone and Togo) the Trace and Maxeigenvalue test statistics indicate the existence of one cointegrating equation. For the remaining 2 countries (Ghana and Niger) the Trace and Max-eigenvalue tests indicate the existence of the 2 cointegrating equations. For the entire ECOWAS region, the statistics tests denote the existence of two cointegrating equations. The existence of cointegration rules out Granger non-causality. 3.3 Testing the short and long run relation 3.3.1. Estimating the Granger Causality The granger causality performed pair-wise for GDP and total energy consumption, income and oil consumption, as well as income and electricity consumption, shows a significant link for almost countries test, at the 5% level of significance (Table 4). For 3 countries, Guinea Bissau, Senegal and Niger, energy consumption, as well as oil and electricity consumption contribute to the economic growth. For Mali and Nigeria, total energy consumption and electricity lead to economic growth (oil consumption does not cause income). For all ECOWAS and Togo, only the total energy consumes causes GDP growth. For Guinea, it s the electricity consumption that causes growth. 5 Verbeek (2001:254) notes that the model with the smallest AIC or SIC is preferred. However, while the two criteria differ in their trade-off between fit and parsimony, the SIC criterion can be preferred.

For, Benin and Sierra Leone, if energy consumption has not cause GDP growth, the GDP growth in contrary lead to the growth of total energy consumption, electricity and oil consumption. For Cap Verde and Guinea, the growth in GDP leads to the growth of total energy consumption but not in the growth of the consumption of oil and electricity. For Ghana and Gambia, the GDP economic growth leads to the growth of total energy and oil consumption. While for Burkina Faso and Senegal, the GDP growth leads to the growth of electricity consumption. 3.3.2. Le Vector Error Correction Model (ECM) The Trace and Max-Eigen statistics indicate the existence of cointegration in the case of 8 countries and the entire ECOWAS. The existence of cointegration requires the Error Correction Model. According to Granger s (1983), if there is a long-run relationship between the variables then there will be a 'short-run' error-correction relationship associated with it. This latter relationship represents an adjustment process by which the deviated actual consumption rate is expected to get back to its long-run equilibrium path. To model this short-run error-correction relationship, the first difference of each variable in the long run relationship should be used, including an error correction term which represents the adjustment of actual to desired equilibrium exchange rate level in the previous period. This error-correction term is expected to be negative sign. Estimating of the ECM equations for all cointegrated countries yields the following results: -Energy consumption has a positive long run impact on growth for Benin, Guinea Bissau, Liberia and Nigeria. And, negative long term impact for the remaining countries (Mali, Niger Sierra Leone and Togo). -Energy consumption seems to have not a long term impact on the Ghana GPD. -Except for Sierra Leone and Guinea Bissau, power consumption has a positive impact on growth for the all the cointegrated countries. -Oil consumption has a positive impact in the long term growth for Benin, Ghana, Sierra Leone, Togo and Niger. And a negative impact for Guinea Bissau, Liberia Mali and Nigeria. The negative impact of oil consumption on growth in Nigeria is unexpected and curious. -For the entire region, there is a long term impact of energy, oil and power consumption on growth. In the short term, except for Benin, Guinea Bissau, Nigeria, Togo and for the ECOWAS, there is a positive relation between energy consumption and growth for the cointegrated countries. The ECM coefficients are in negative sign for Ghana, Liberia, Mali, Niger and Sierra Leone. There is short run elasticity between energy consumption and growth for these countries. Short run elasticity exists also for electricity consumption and growth for Sierra Leone and the ECOWAS. For the oil consumption and GDP, this short run elasticity exists for Ghana, Liberia, Mali and Sierra Leone. Short run elasticity coefficient exists also between current GDP and the past income except for Liberia.

-Current GDP thus depends as much on current energy, electricity, oil consumption as on past GDP for Sierra Leone. -Current GDP depends on past GDP and energy consumption for: Ghana and Niger. -Current GDP depends on past GDP and energy consumption and oil consumption for Mali -For Liberia current GDP depends on energy consumption and oil consumption. -For the ECOWAS, current GDP depends on only past GDP and the electricity consumption. IV. Conclusions and policy implications In this paper a recently developed cointegration test and a modified version of the Granger causality test were applied to investigate the long run and causal relationship between GDP and energy use for 15 West Africa countries from 1980 to 2008 to re-examine the co-movement and causal relationship between GDP and energy consumption according to the short-run and the long-run dynamics of energy consumption and GDP, we refute the neutrality hypothesis advanced in respect of developing countries for the energy income relationship. Our evidence shows results suggesting that there is a long-run steady-state relationship between energy consumption and GDP for a cross-section of countries after allowing for a country-specific effect. Energy consumption is found to Granger cause GDP or vice versa for the many countries of the panel. The implication of this finding is that a high level of economic growth leads to high level of energy demand and vice-versa. The results of a unidirectional long-run causal relationship and a uni-directional short-run causal relationship running from energy to GDP show that energy consumption leads economic growth. This implies that energy consumption bears the burden of the short-run adjustments to reestablish the long-run equilibrium. In other words, high energy consumption tends to have high economic growth. Current as well as past changes in energy consumption have a significant impact on a change in income in some ECOWAS countries. For these countries, energy is an important factor for economic development and energy conservation may harm economic growth regardless of it being transitory or permanent. This direction of causation expounds future energy use concerning environmental protection and economic development. The reverse effect is observed for six of the 15 countries. For these countries, economic poverty leads to energy poverty. GDP growth has a positive impact on energy consumption. The results show in some case that from the lagged dynamic terms, the short run changes in economic growth are in part responsible for future

changes in energy consumption. That is, a faster rate of growth promotes higher energy consumption. Also, the variable has a crucial impact on growth through the adjustment of error correction terms (ECM), which are significant and have a correct sign for many of the 15 countries. The ECMs in Table 5 indicate that there exists a mechanism in correcting the disequilibrium between economic growth and energy consumption. The implication of this finding is that reducing income could lead to a fall in energy consumption. This means that investment and other efficient measures that increase energy supply can be implemented, but such measures should not be at the expense of the environment. The directional causality running from economic growth to energy consumption may statistically suggest that energy consumption measures may be taken without jeopardizing economic development. This is not to suggest however, that energy consumption level should be reduced. The option therefore might be for these countries to enhance the level of efficiency Indeed, in order not to adversely affect economic growth, energy conservation policies that aim at reducing energy must rather find ways of reducing consumer demand. This sort of policy could be achieved through an appropriate mix of energy taxes and subsides. This paper did not provide a definite stand on the existence or non-existence a long run nor a causal relationship between energy consumption and economic growth; but our results show the following: there was evidence of a long run relationship for only four of the 15 countries and causality for 14 countries. Broadly speaking, our results are as mixed as the other results in the literature: (1) past values of economic growth have a predictive ability in determining present values of energy consumption in some countries, (2) past values of energy consumption have a predictive ability in determining present values of economic growth; (3) there was feedback in some ECOWAS countries and (4) there was a lack of causal relationship for other countries.

Appendixes Table 2: testing for non-stationarity (t-statistics for the Phillips-perron test, Newey-West bandwith using kernel) Variable Series first -differences Intercept Trend Lag truncation Stationarity level Benin Gdp -5.180005* Yes No 3 I(1) totalener -4.780641* Yes No 3 I(1) Elec -14.74944* Yes No 3 I(1) petrol -4.514038* No No 3 I(1) Burkina Faso Gdp -5.709245* Yes No 3 I(1) totalener -4.533181* Yes No 3 I(1) Elec -4.535194* Yes No 3 I(1) petrol -4.325932* Yes No 3 I(1) Cap Verde Gdp -4.661351* Yes No 3 I(1) totalener -5.021421* No No 3 I(1) Elec -6.824852* Yes No 3 I(1) petrol -5.130248* No No 3 I(1) Côte d Ivoire Gdp -2.886189* Yes No 3 I(1) totalener -7.789501* Yes No 3 I(1) Elec -6.392417* Yes No 3 I(1) petrol -7.748150* No No 3 I(1) Gambia Gdp -5.996080* Yes No 3 I(1) totalener -5.114299* Yes No 3 I(1) elec -6.252657* Yes No 3 I(1) petrol -3.923676* No No 3 I(1) Ghana Gdp -2.914273*** Yes No 3 I(1) totalener -4.087595* Yes No 3 I(1) elec -5.872009* Yes No 3 I(1) petrol -4.955379* No No 3 I(1) Guinea Elec -4.315765* Yes No 3 I(1) petrol -6.505641* No No 3 I(1) Gdp -3.955406* Yes No 3 I(1) totalener --7.455777* No No 3 I(1) Guinea bissau Elec -4.437403* No No 3 I(1) petrol -4.363359* Yes No 3 I(1) Gdp -5.944460* No No 3 I(1) totalener -4.739107* Yes No 3 I(1) Liberia Gdp -2.789824* No No 3 I(1) totalener -3.200977* No No 3 I(1) elec -4.948056* No No 3 I(1) petrol -4.376843* No No 3 I(1) Mali Gdp -5.996080* Yes No 3 I(1) totalener -5.114299* Yes No 3 I(1) elec -6.252657* Yes No 3 I(1) petrol -3.923676* No No 3 I(1) Niger Gdp -4.263920* No No 3 I(1) totalener -8.747392* No No 3 I(1) Elec -4.791049* Yes No 3 I(1) petrol -4.269998* No No 3 I(1) Nigeria Gdp -5.147297* Yes No 3 I(1) totalener -6.508982* Yes No 3 I(1) Elec -8.169327* Yes No 3 I(1) petrol -4.269998* Yes No 3 I(1) Senegal Gdp -6.030870* Yes No 3 I(1) totalener -5.084853* Yes No 3 I(1) elec -7.290296* Yes No 3 I(1) petrol -4.247305* No No 3 I(1) Sierra Leone Gdp -3.897280* No No 3 I(1) totalener -5.104250* No No 3 I(1) elec -5.041729* No No 3 I(1) petrol -4.798560* No No 3 I(1) Togo Gdp -4.271012* No No 3 I(1) totalener -6.285059* Yes No 3 I(1) Elec -6.884989* Yes No 3 I(1) ECOWAS petrol -6.286521* No No 3 I(1) Gdp -6.645795* Yes No 3 I(1) totalener -7.116333* Yes No 3 I(1) Elec -6.457060* Yes No 3 I(1) petrol -4.089724* Yes No 3 I(1) The signs (*), (**),(***) show the null hypothesis is confirmed at the 1%, 5% and 10% level at significantce respectively. Source: author calculation with EViews.

Table 3: Trace and Max-Eigen tests results Null hypothesis Alternative hypothesis Trace statistic Max-Eigen stsatitic Series: GDP ENERPRICE TOTALENER ELEC PETROL Benin r = 0 r> 1 86.50798* 50.35688* r 1 r >2 36.15110 16.16160 r 2 r >3 19.98950 14.52439 r 3 r >4 5.465106 5.071977 Ghana r = 0 r> 1 152.0542* 79.82786* r 1 r >2 72.22635* 49.83258* r 2 r >3 22.39377 17.99661 r 3 r >4 4.397155 3.087819 Guinea Bissau r = 0 r> 1 117.6651* 64.54279* r 1 r >2 53.12231* 24.74505 r 2 r >3 28.37725 15.03962 r 3 r >4 13.33764 10.71987 Liberia r = 0 r> 1 69.28356* 35.11602* r 1 r >2 34.16754 19.42944 r 2 r >3 14.73810 11.41836 r 3 r >4 3.319736 3.243478 Mali r = 0 r> 1 78.97551* 32.51905 r 1 r >2 46.45646 23.56874 r 2 r >3 22.88773 12.25129 r 3 r >4 10.63644 9.263213 r 4 r> 1 1.373226 1.373226 Niger r = 0 r> 1 76.99944* 50.21058* r 1 r >2 48.55865* 44.40055* r 2 r >3 23.71239 11.59795 r 3 r >4 5.742298 7.328168 Nigeria r = 0 r> 1 59.71565* 34.30672* r 1 r >2 25.40892 14.68273 r 2 r >3 10.72619 10.67182 r 3 r >4 0.054372 0.054372 Sierra leone r = 0 r> 1 75.12847* 30.86246 r 1 r >2 44.26601 19.62770 r 2 r >3 24.63831 16.52757 r 3 r >4 8.110740 8.042337 r 4 r> 1 0.068404 0.068404 Togo r = 0 r> 1 77.05524* 32.49576 r 1 r >2 44.55948 21.69175 r 2 r >3 22.86773 14.06294 r 3 r >4 8.804790 7.416298 ECOWAS r = 0 r> 1 132.4446* 54.91331* r 1 r >2 77.53133* 52.76021* r 2 r >3 24.77111 11.03258 r 3 r >4 13.73854 10.15686 * denotes rejection of the hypothesis at the 5% level 14

Table 4: results of Granger causality (the table reports the existence of pair-wise granger causality as well as the relevant F-statistics, the critical value at the 5% level being 2,99, results significant at the 5% level are in bold. GDP ENERGY OIL ELECTRICITY BENIN GDP n.a Yes (5.00685) Yes (28.3965) Yes (5.66807) Energy No (1.16949) Oil No (0.54083) Electricity No (2.85530) BURKINA FASO No (0.00134) GDP n.a No (3.26655) No (2.40970) Yes (8.61987) Energy No (0.58368) Oil No (2.89547) Electricity No (0.09602) CAP VERDE GDP n.a Yes (4.81259) No (3.13541) No (1.95647) Energy No (0.02691) Oil No (0.19069) Electricity No (2.16223) COTE D IVOIRE GDP n.a No (3.74629) No (2.79142) No (2.57011) Energy No (0.03258) Oil No (5.24724) Electricity No (1.00053) GAMBIA GDP n.a Yes (10.5826) Yes (21.5880) No (2.03192) Energy No (1.63902) Oil Yes (19.2530) Electricity No (2.03192) GHANA GDP n.a Yes (5.11637) Yes (5.97391) No (1.09524) Energy No (1.23587) Oil No (1.52324) Electricity No (0.00465) GUINEA GDP n.a Yes (10.0070) No (1.1687) No (0.93944) Energy No (0.10901) Oil No (3.88885) Electricity Yes (6.79698) GUINEA BISSAU GDP n.a No (0.22912) No (1.82298) No (0.21449) Energy Yes (5.27223) Oil Yes (6.06513) Electricity Yes (6.23629) LIBERIA GDP n.a No (0.62901) No (0.00035) No (0.14325) Energy No (0.00486) Oil No (1.59298) Electricity No (0.14736) MALI GDP n.a No (0.07073) No(0.11556) No (1.75452) Energy Yes (8.06607) Oil No (2.95648) Electricity yes (13.2753) NIGER GDP n.a No (8.58173) No (6.42025) No (2.67823) Energy Yes (3.34748) Oil Yes (0.07864) Electricity Yes (3.65205) NIGERIA GDP n.a No (1.79060) No ( 3.28986) No (3.09697) Energy Yes (23.2239) Oil No (0.22874) Electricity Yes( 5.76520) SENEGAL GDP n.a No (1.13025) No (1.04348) Yes (10.8327) 15

Energy Yes (6.76678) Oil Yes (4.41633) Electricity Yes (7.28986) SIERRA LEONE GDP n.a Yes (4.47986) Yes (5.61159) Yes (4.47649) Energy No (0.07314) Oil No (0.51545) Electricity No (0.04421) TOGO GDP n.a No (0.00892) No (5.01607) No (0.40324) Energy Yes (9.44478) Oil No ( 0.43248) Electricity No (3.50247) ECOWAS GDP n.a No (0.70450) Yes (7.67805) No (3.58146) Energy Yes (45.1209) Oil No (1.99096) Electricity No (3.42823) Table 5: Cointegrating vector and speed of adjustment parameters for cointegrated systems Cointegrating Equation Error correction C GDP(-1) ENER(-1) ELEC(-1) OIL(-1) D(GDP) D(ENER) D(ELEC) D(OIL) Benin - 13.7175 1.0000-0.2947 (0.1046) [-2.8167] -0.4615 (0.0331) [-13.933] -0.0438 (0.0123) [-3.5514] -0.1350 (0.0724) [-1.863] 0.05238 (0.1596) [0.3282] 0.17477 (0.1762) [0.9921] 1.01305 (0.8426) [1.2023] Ghana Guinea Bissau -3.6777 1.0000 1.0000-20.1299 1.0000-3.1823 (1.3614) [-2.3375] -0.3769 (0.0358) [-10.543] 0.32691 (0.0641) [5.10034] -0.49160 (0.0311) [-15.8406] 1.25601 (0.50493) [2.4875] -0.005380 (0.04492) [-0.1198] -0.09140 (0.1082) [-0.8450] -0.109689 (0.05639) [-1.9451] 0.02792 (0.02123) [1.3150] 3.96625 (0.51061) [7.76762] 0.00205 (0.20512) [0.00998] -0.34503 (0.16939) [-2.03689] 0.07951 (0.0556) [1.4314] Liberia 12.1412 1.0000-1.2789 (0.6070) [-2.107] -2.08877 (0.4126) [-5.0628] 1.19218 (0.2581) [4.6185] 0.11644 (0.08799) [1.3233] -0.03368 (0.03638) [-0.9256] 0.01015 (0.04453) [0.2280] -0.1680 (0.0914) [-1.8390] Mali - 24.0760 1.0000 0.30313 (0.7299) [0.4153] -0.44374 (0.1465) [-3.0299] 0.24896 (0.17549) [1.4186] -0.20470 (0.0766) [-2.6716] -0.02744 (0.04679) [-0.58648] 0.25040 (0.15401) [1.6258] -0.425812 (0.1697) [-2.5099] Niger - 13.8530 1.0000 0.1837 (0.12505) [1.46903] -1.1467 (0.14501) [-7.9072] -0.15652 (0.04475) [-3.4980] -0.18047 (0.1354) [-1.3329] -0.19237 (0.32845) [-0.5857] 0.37325 (0.11690) [3.19291] 0.02270 (0.34497) [0.06579] Nigeria Serra Leone - 54.4346 0.14952 1.0000 1.0000-1.27370 (0.2029) [-6.2782] 0.20222 (2.93482) [0.06890] -0.34076 (0.0969) [-3.5176] 3.4448 (1.2773) [2.6970] 0.10755 (0.11672) [0.9214] -2.74557 (0.9519) [-2.8843] -0.30245 (0.0567) [-5.3356] -0.02178 (0.0119) [-1.8237] 0.03465 (0.0208) [1.6668] -0.00829 (0.00435) [-1.9041] 0.13334 (0.2714) [0.4913] -0.02240 (0.0195) [-1.1483] 0.20888 (0.18704) [1.1167] -0.00612 (0.0237) [-0.2580] Togo -33.262 1.0000 3.0858 (0.5355) [5.7626] -2.2019 (0.3405) [-6.4667] -0.54759 (0.1318) [-4.1535] -0.00223 (0.0537) [-0.0416] 0.00631 (0.0540) [0.1168] 0.18014 (0.0844) [ 2.1341] 0.27415 (0.2838) [0.9661] ECOWAS 0.8437 1.0000-0.2391 (0.0998) [-2.3955] -0.01521 (0.0183) [-0.8334] -0.45015 (0.0670) [-6.7144] -0.01521 (0.0183) [-0.8334] 0.01906 (0.0277) [0.6894] -0.09710 (0.2502) [-0.3881] 0.0263 (1.6068) [0.0164] Source: author calculation with Eviews. 16

References Akara, A., Long, T., 1980. On the relationship between energy and GNP: a re-examination. Journal of Energy Development 5, 326 331. Asafu-Adjaye, J., 2000. The relationship between energy consumption, energy prices and economic growth: time series evidence from Asian developing countries. Energy Economics, 615 625. Granger, C.W.J., 1988. Causality, cointegration and control. Journal of Economic Dynamics and Control 12, 551 559. Besant-Jones, J.E., 2006. Reforming Power Markets in Developing Countries: What Have We Learned? Paper no. 19, Energy and Mining Sector Board, The World Bank Group, Washington, DC. Chevalier, J-M. and Ouedraogo, N. (2009). Energy poverty and economic development, In The New Energy Crisis: Climate, Economics, Geopolitics, Eds Palgrave, 2009. Engle, R.F., Granger, C.W.J., (1987). Co-integration and error correction: representation, estimation and testing. Econometrica, Vol 55, n 2, March, pp 251-276. Granger, C.W.J. (1969). Investigating causal relation by econometric and cross-sectional method. Econometrica 37, 424 438. Granger, C.W.J., (1986). Developments in the study of cointegrated variables. Oxford Bulletin of Economics and Statistics. Vol 48. pp 213-228. Johansen, S., Juselius, K., (1990). Maximum likelihood estimation and inferences on cointegration with application to the demand for money. Oxford Bulletin of Economics and Statistics. Vol 52, pp 160-210. Hadri, K., 2000. Testing for stationarity in heterogeneous panel application. Econometric Journal 3, 148 161. Johansen, S., 1988. Statistical analysis of cointegration vectors. Journal of Economics Dynamic and Control 12, 231 254. Keppler, J., Bourbonnais, R. and Girod, J. (with an introduction by Chevalier J-M) (2006). The Econometrics of Energy Systems. Editions Palgrave Macmillan, Décembre 2006. Kraft, J., Kraft, A., 1978. On the relationship between energy and GNP. Journal of Energy and Development 3, 401 403. Lee, C.C. (2005). Energy consumption and GDP in developing countries: A cointegrated panel analysis. Energy Economics 27, 415 427. Lee, C.C., Chang, C.P. (2007) Energy consumption and GDP revisited: a panel analysis of developed and developing countries, Energy Economics, 29, 1206-1223. Mahadevan, R., Asafu-Adjaye, J., 2007. "Energy consumption, economic growth and prices: A reassessment using panel VECM for developed and developing countries," Energy Policy, 35, 2481-2490. Masih, A.M.M., Masih, R., 1996. Energy consumption, real income and temporal causality: results from a multicountry study based on cointegration and error-correction modelling techniques. Energy Economics 18, 165 183. Nachane, D.M., Nadkarni, R.M., Karnik, A.V., 1988. Co-integration and causality testing of the energy-gdp relationship: a crosscountry study. Applied Economics 20, 1511 1531. 17

Pineau, P.-O., 2007. Electricity sector integration in West Africa. Energy Policy 36 210 223 Wolde-Rufael, Y., 2004. Disaggregated energy consumption and GDP, the experience of Shangai, 1952 1999. Energy Economics 26, 69 75. 18