The impact of energy demand on economic growth: A new empirical evidence for Madagascar Franck Ramaharo*1, Yves Razanajatovo1, Fabienne Ravelomanantsoa2, Melodia Ramarosandratana2, and Emanuella Aljaona3 1Economic Modeling Unit, Ministry of Economy and Finance 2Department of Economics, Catholic University of Madagascar 3Department of Economics, University of Antananarivo Antananarivo 101, Madagascar January 7, 2024 Abstract This study investigates the impact of energy demand on Madagascar’s economic growth from 2007Q1 to 2022Q4. Drawing upon a rich dataset from Malagasy sources, we applied the ARDL bounds testing approach and found cointegration among the series. We found that, in the long run, electricity and petroleum consumption have positive significant effects on economic growth, while energy imports and global prices have negative significant effects. We further applied Granger- causality test based on Error Correction Model to examine causal relationships. The results re- vealed that in the short run, there are unidirectional causal effects running from electricity con- sumption, energy imports, and global prices to economic growth. The test also revealed that both energy demand and global prices have a long-run causal effect on economic growth. Our findings confirms that Madagascar is an energy-dependent economy, and provide valuable insights for poli- cymakers to design effective energy policies that promote economic growth and energy security. Keywords: energy demand, electricity consumption, petroleum consumption, energy imports, economic growth, Madagascar Disclaimer. The views presented in this working paper are solely those of the authors and do not necessarily reflect the positions of their affiliations or any other mentioned entities. 1 Introduction Statistical data related to energy in Madagascar generally focuses on electricity and petroleum (IN- STAT, 2022b). While the “Energy and Growth” theme is popular among Malagasy economists and scholars, research addressing the impact of energy demand on economic growth within Madagascar re- mains scarce. This scarcity extends to studies utilizing primary data from Malagasy authorities, likely due to limited public access. To the best of our knowledge, only a handful of publications like Vonini- rina & Andriambelosoa (2014) and Andriamanga (2017) have empirically examined this relationship using Malagasy data. Recognizing this gap, our study presents a renewed empirical assessment of the impact of energy demand on economic growth in Madagascar. The originality of this study lies in its exclusive use of primary data from relevant Malagasy authorities. We apply the Autoregressive *Corresponding author: franck.ramaharo@gmail.com 1 mailto:franck.ramaharo@gmail.com Distributed Lag (ARDL) bounds testing approach to comprehensively analyze the dynamic relation- ship between energy consumption and economic growth in Madagascar based on quarterly data from 2007Q1 to 2022Q4. Additionally, we employ Granger-causality tests based on an error correction model to comprehensively reassess the causal relationship between these variables, previously explored by Voninirina & Andriambelosoa (2014) and Andriamanga (2017). 2 Brief empirical review Initially, Voninirina & Andriambelosoa (2014) conducted an econometric analysis to investigate the relationship between energy consumption and economic growth in Madagascar. Their research aimed to discern the existence of a long-term relationship between economic growth and energy consumption, using annual data from 1980 to 2011 (retrieved from the World Bank’s World Development Indicators and JIRAMA) and applying cointegration methodology. They used Granger’s method to determine the direction of the relationship during the same period. Their findings revealed the lack of a long- term relationship between the variables, however, they identified a bidirectional causality relationship. Later, Andriamanga (2017) investigated the empirical relationship between energy consumption and economic growth in Madagascar for the 1995–2015 period. The data sources include annual statistics from the World Bank, and the former Ministry of Water, Energy, and Hydrocarbons. The empirical methodology was conducted through the Error Correction Model (ECM) framework to analyze the short and long-term dynamics. In contrast to Voninirina & Andriambelosoa (2014)’s results, Andriamanga (2017)’s findings revealed a significant long-term relationship between energy consumption (electricity and petroleum) and economic growth. Additionally, bidirectional Granger causality between petroleum consumption and economic growth was identified, as well as unidirectional Granger causality running from electricity consumption to economic growth. It is also worth noting the empirical results which were conducted in the context of panel data anal- ysis. Nondo et al. (2010) conducted panel cointegration techniques and panel causality tests on a group of 19 African countries within the Common Market for Eastern and Southern Africa (COMESA). The tests, based on data spanning from 1980 to 2005, revealed that energy consumption (proxied by BTU of energy) stimulates GDP growth, in the short and long run for low-income countries in COMESA. In contrast, Kahsai et al. (2012) tested the relationship between energy consumption and economic growth in Sub-Saharan African countries, using a panel co-integration approach for 40 countries for the 1980–2007 period, and their findings support the neutrality hypothesis in the short-run, i.e., no causality between energy consumption and economic growth in the short-run. However, they find a bidirectional interdependence of energy consumption and economic growth in Sub-Saharan Africa in the long run. In addition, Karanfil & Li (2015) used a panel of 160 countries for the 1980–2010 period to examine the long- and short-run dynamics between electricity consumption and economic activities. Specifically, they used panel cointegration techniques to estimate the relationship between electricity consumption, real GDP, electricity net imports, and urbanization ratio. They identified a bidirectional short-run causality between electricity consumption and real GDP for low-income coun- tries. Later, Atems & Hotaling (2018) assessed the impact of electricity generation instead of electricity consumption, considering the fact that not all of the electricity generated in a country is eventually consumed because of transmission and distribution losses, as well as theft. Therefore, they investi- gated the impact of total electricity generation on economic growth, disentangled into renewable and non-renewable effects, using data from a panel of 174 countries over the period 1980 to 2012, and using the System Generalized Methods of Moments approach. Their results indicate a positive relationship between renewable and nonrenewable electricity generation and economic growth. Similarly, Espoir et al. (2023) investigated the impact of renewable electricity consumption (REC) and non-renewable electricity consumption (NREC) on economic growth in 51 African countries between 1980 and 2018. They found that NREC has a positive impact on economic growth but surprisingly, the effect of REC is 2 negative in the COMESA region. Moreover, they find that both REC and NREC have positive effects on growth in the Southern African Development Community (SADC) region. Recently, Mmbaga et al. (2023) examined the relationship between energy consumption and economic growth in the East African sub-region (2012-2021). They employed panel data from 13 countries and used Pooled Ordi- nary Least Squares, Fixed Effects, and Generalized Method of Moments. Their findings suggest that both higher energy prices and increased energy consumption reduce economic growth, highlighting the need for careful energy management in this developing region. 3 Data and methodology 3.1 Data descriptions Data used in this study are mainly obtained from the National Institute of Statistics (INSTAT), the Malagasy Office of Hydrocarbons (OMH), the Jiro sy Rano Malagasy (JIRAMA)–the national water and electricity company of Madagascar, and the General Directorate of Customs (Douane Malagasy) within the Ministry of Economy and Finance. Our empirical estimation is based on quarterly observations that cover the period 2007Q1 to 2022Q4. Data are summarized in Table 1. Table 1: Variable definition. Variables Notation Description (unit) Source Economic growth GDP Real gross domestic product (billions of Ariary, at 2007 constant prices) INSTAT Electricity consumption ELEC Sum of low and medium voltage electric- ity consumption (GWh) JIRAMA; INSTAT Petroleum consumption PETRC Sum of naphtha, jet fuel, aviation fuel, liquefied petroleum gas, kerosene, super- unleaded petrol, gasoline, gas oil, and heavy fuel oil consumption (cubic meter) OMH; INSTAT Energy imports ENRMG Sum of imports of petroleum products and other fuel products (% total volume of imports of goods) Douane Malagasy; INSTAT Global prices CPI Consumer price index (2016= 100) INSTAT The following descriptions provide a concise overview of the variables used in our analysis. Unless otherwise indicated, the reported shares represent averages from 2007Q1 to 2022Q4. • Madagascar’s quarterly GDP at constant 2007 prices are taken from the INSTAT quarterly na- tional accounts which provide information consistent with annual national accounts. The quar- terly GDP allows better responsiveness to economic shocks and for short-term monitoring of economic policies (INSTAT, 2019a). As of the 2007 base year, Madagascar’s GDP splits across primary (28.1%), secondary (10.9%), and tertiary (54.5%, including Financial intermediation services indirectly measured) sectors, with the remaining 6.5% attributed to indirect taxes (IN- STAT, 2019b). • The JIRAMA manages the distribution of electricity at two distinct voltage levels. Low voltage caters to the needs of households and public lighting (61.6%), while medium voltage serves the demands of the three major economic sectors: primary (0.3%), secondary (24.9%), and tertiary (13.2%). As of 2021, 45.2% of the JIRAMA’s electricity is generated from renewable sources (95.6% of which comes from hydropower plants), with the remaining 54.8% generated by thermal power plants running on heavy fuel oil and gas oil (ARCEB, 2022, p. 53). 3 • According to data from the petroleum market monitored by the OMH, petroleum products consist of naphtha, liquefied petroleum gas (LPG), aviation fuel, jet fuel, super-unleaded petrol, gaso- line, kerosene, gas oil, and heavy fuel oil. These products are distributed across various sectors of activity, including transportation, aviation, bunkers (maritime operations), the JIRAMA, Am- batovy and QIT Madagascar Minerals (QMM)–two of Madagascar’s major mining companies– households and industries, and construction projects (OMH, 2023a). Based on available data ranging from 2016 to 2022, transportation has the largest share among all sectors (about 50.0%), primarily consuming gas oil and super-unleaded petrol (OMH, 2023c). The second largest con- sumer is the JIRAMA, accounting for roughly 21.8% of petroleum products. Households and industries together claim approximately 10% of the share. Households primarily utilize LPG for cooking, and kerosene for lighting and cooking, especially in rural areas (Naidoo & Loots, 2020). Meanwhile, industries rely on gasoline for vehicles, gas oil for generators and machinery, and heavy fuel oil for specific processes. • Madagascar’s imports of goods can be categorized into five main groups: food, energy, equip- ment, raw materials, and miscellaneous items. Raw materials represent the dominant category, accounting for nearly half of all imports. Energy follows closely, constituting a significant quarter of the total volume imports of goods. Within the energy category, according to the Harmonized System (HS) classification, coal and petroleum oils are the primary imports. For example, as of 2022, the main energy imports consist of selected sub-items for the following classification. – HS2701 Coal; briquettes, ovoids and similar solid fuels manufactured from coal. – HS2703 Peat, incl. peat litter, whether or not agglomerated. – HS2704 Coke and semi-coke of coal, of lignite or of peat, whether or not agglomerated; retort carbon. – HS2707 Oils and other products of the distillation of high temperature coal tar; similar products in which the weight of the aromatic constituents exceeds that of the non-aromatic constituents. – HS2710 Petroleum oils and oils obtained from bituminous minerals (excl. crude); prepara- tions containing >= 70% by weight of petroleum oils or of oils obtained from bituminous minerals, these oils being the basic constituents of the preparations, n.e.s.; waste oils con- taining mainly petroleum or bituminous minerals. – HS2711 Petroleum gas and other gaseous hydrocarbons. – HS2713 Petroleum coke, petroleum bitumen and other residues of petroleum oil or of oil obtained from bituminous minerals, n.e.s. – HS4401 Fuel wood, in logs, billets, twigs, faggots or similar forms; wood in chips or parti- cles; sawdust and wood waste and scrap, whether or not agglomerated in logs, briquettes, pellets or similar forms. – HS4402 Wood charcoal, incl. shell or nut charcoal, whether or not agglomerated (excl. wood charcoal used as a medicament, charcoal mixed with incense, activated charcoal and char- coal in the form of crayons). Note that data are sourced from the Douane Malagasy (2023) and consolidated with that of the International Trade Centre (ITC)’s Trade Map (https://www.trademap.org/). • Consumer Price Index (CPI) data in Madagascar, compiled by INSTAT, uses the 2016 reference year as its basis and is referred to as the new CPI (INSTAT, 2019c). Within this new CPI, energy products hold a 7.8% share of the overall index. However, when categorized by function, the im- pact of energy is even more pronounced. Housing, water, electricity, gas, and other combustibles occupy a 20.3% share of the total CPI, suggesting that energy costs contribute to a substantial portion of household expenses in Madagascar (AfDB, 2023). 4 https://www.trademap.org/ 3.2 Empirical methodology In this study, we applied the TRAMO/SEATS procedure (Gómez & Maravall, 1997) to remove seasonal and random fluctuations from the initial series of data. We then take GDP, ELEC, PETRC, and ENRMG variables in their natural logarithmic form. Next, we conduct unit root tests to justify that the variables are stationary at level, i.e., I(0), or at the first difference, i.e., I(1). We check stationarity through the Augmented Dickey-Fuller (ADF), Philips-Perron (PP), and Kwiatkowski-Phillips-Schmidt- Shin (KPSS) unit root tests (Dickey & Fuller, 1979; Phillips & Perron, 1988; Kwiatkowski et al., 1992). After determining the stationarity of the variables, we used the Autoregressive Distributed Lag (ARDL) bounds testing approach developed by Pesaran & Shin (1999) and Pesaran et al. (2001) to examine the impact of energy demand on economic growth in Madagascar. We build our model based on the availability of data and the empirical works of Alnour (2021), Choi & Yoo (2016), Gbadebo & Okonkwo (2009), Ghosh (2009), Kwakwa (2012), Murshed et al. (2020), Narayan & Singh (2007), Ngoc (2019), Odhiambo (2009), Okey (2009), Onakoya et al. (2013), Sama & Tah (2016). The functional relationship of our model, which we assume linear, then takes the following form: GDPt = f (ELECt,PETRCt,ENRMGt,CPIt), (1) where variables are defined in Table 1. In (1), ENRMG is considered as a proxy for Madagascar’s energy dependence, and CPI serves as a price level index, reflecting changing energy prices and in- tegrating them into the dynamic relationship between energy consumption and its economic impact (Hondroyiannis et al., 2002; Kahsai et al., 2012; Mmbaga et al., 2023). The ARDL representation of (1), denoted ARDL(p1, p2, p3, p4, p5), takes the following form: GDPt =β0+ p1∑ i=1 β1,iGDPt−i+ p2∑ i=0 β2,iELECt−i+ p3∑ i=0 β3,iPETRCt−i+ p4∑ i=0 β4,iENRMGt−i+ p5∑ i=0 β5,iCPIt−i+εt, where (p1, p2, p3, p4, p5) are the optimal lag length of the respective variables which are determined by the Hannan–Quinn information criterion (HQC), β0 is a constant, the β j,i ’s represent the coefficients of the model, and εt is the usual white noise residuals. The ARDL bounds testing approach is applicable irrespective of whether the variables are I(0) or I(1). It is appropriate for estimating both the short- and long-run coefficients, and selecting different lag lengths to address the endogeneity and serial correlation problems. Moreover, it encompasses the estimation of an unrestricted error-correction model which can be expressed as ∆GDPt =β0 + p1−1∑ i=1 β′ 1,i∆GDPt−i + p2−1∑ i=0 β′ 2,i∆ELECt−i + p3−1∑ i=0 β′ 3,i∆PETRCt−i + p4−1∑ i=0 β′ 4,i∆ENRMGt−i + p5−1∑ i=0 β′ 5,i∆CPIt−i +α1GDPt−1 +α2ELECt−1 +α3PETRCt−1 +α4ENRMGt−1 +α5CPIt−1 +εt, (2) where ∆ denotes the first-difference operator, and the β′ coefficients represent the short-run relation- ships, while the α coefficientd represent the long-run relationships. To examine a potential cointe- grated relationship between variables, the null hypothesis of no cointegration (H0) : α1 =α2 =α3 =α4 =α5 = 0, is tested against the alternative hypothesis of cointegration (H1) : α1 ̸=α2 ̸=α3 ̸=α4 ̸=α5 ̸= 0 through the F-statistic (Wald test). The F-statistic obtained from the bounds test is usually compared with the lower and upper bound values proposed by Pesaran et al. (2001). However, in the case of a 5 small sample, as is our case, we use Narayan (2005)’s critical values to determine any long-term asso- ciation among the variables. The lower critical bound assumes all the variables are I(0), meaning that there is no cointegration relationship between the examined variables. The upper bound assumes that all the variables are I(1), meaning that there is cointegration among the variables. If the computed F- statistic is greater than the upper bound critical value, then (H0) is rejected. If the F-statistic is below the lower bound critical value, then (H0) cannot be rejected. If the computed F-statistics falls between the lower and upper bound, then the results are inconclusive. If the variables are cointegrated, the long-run coefficients of each variable can be estimated through an error correction model: ∆GDPt =β0 + p1−1∑ i=1 β′ 1,i∆GDPt−i + p2−1∑ i=0 β′ 2,i∆ELECt−i + p3−1∑ i=0 β′ 3,i∆PETRCt−i + p4−1∑ i=0 β′ 4,i∆CPIt−i + p5−1∑ i=0 β′ 5,i∆ENRMGt−i +ϕECTt−1 +εt, where ECTt−1 stands for the one-period lagged error-correction term, which is the residuals that are obtained from the estimated cointegration model of (2), and ϕ denotes the measure for the speed of adjustment at which any distortion from the long-run equilibrium in the previous period is restored in the current period. The stability of the ARDL estimates is then evaluated using a set of diagnostic tests. The normal- ity of the residuals is evaluated using the Jarque-Bera test. The Breusch-Godfrey Lagrange multiplier (LM) test is used to explore the serial correlation problems. We diagnose the heteroskedasticity issues with the Breusch-Pagan-Godfrey heteroskedasticity and Autoregressive Conditional Heteroskedastic- ity (ARCH) tests. The correct functional form of the models is assessed using the Ramsey Regression Specification Error Test (RESET). Finally, the structural stability of the ARDL estimates is checked using the cumulative sum of recursive residuals (CUSUM) and the cumulative sum of squares of re- cursive residuals (CUSUMSQ) tests advanced by Brown et al. (1975). Furthermore, we used the Granger-causality method to explore the causal interaction among the investigated variables. To achieve this, we run the pairwise Granger-causality tests and the Vector Er- ror Correction Model (VECM) for the short- and long-run relationships, respectively (Engle & Granger, 1987). The estimate of the dynamic VECM is given as follows: ∆Yt =B0 + ℓ∑ i=1 Bi∆Yt−i +ΦECTt−1 +Λt, (3) with Yt =  GDPt ELECt PETRCt ENRMGt CPIt  , B0 =  β0,1 β0,2 β0,3 β0,4 β0,5  , Bi =  β1,1,i β1,2,i β1,3,i β1,4,i β1,5,i β2,1,i β2,2,i β2,3,i β2,4,i β2,5,i β3,1,i β3,2,i β3,3,i β3,4,i β3,5,i β4,1,i β4,2,i β4,3,i β4,4,i β4,5,i β5,1,i β5,2,i β5,3,i β5,4,i β5,5,i  , Φ=  ϕ1 ϕ2 ϕ3 ϕ4 ϕ5  , Λt =  ε1,t ε2,t ε3,t ε4,t ε5,t  , where B0 is a constant vector, Bi is a matrix of the coefficients that need to be estimated and which will demonstrate the short-run causality across the variables, Φ is a vector of the coefficients of the one-period lagged error-correction term (ECTt−1) which indicates the adjustment of the dependent variables to their long-run equilibrium, and Λt is a vector of the white noise residuals. The optimal lag length ℓ in (3) is selected using the Schwarz’s Bayesian information criterion (BIC). The advantage of using an error-correction model to test for causality is that it allows testing for short-run causality through the lagged difference explanatory variables and for long-run causality through the lagged error-correction term. On the one hand, short-run Granger-causality is examined by testing (H0) : βi, j,1 =βi, j,2 =βi, j,3 = ·· · =βi, j,ℓ = 0, using Wald test’s F-statistics, where i, j = 1, . . . ,5, 6 with i ̸= j. For example, β1,2,1 = β1,2,2 = β1,2,3 = ·· · = β1,2,ℓ = 0 indicates that electricity consumption does not Granger-cause economic growth. On the other hand, long-run Granger-causality is examined by testing (H0) :ϕk = 0, using t-statistics, where k = 1,2,3,4,5. 4 Empirical results 4.1 Unit root tests Before conducting the ARDL bounds test, it is crucial to test for variable stationarity to ensure they are either I(0) or I(1). Table 2 presents the findings of our unit root tests. We examined stationarity with both intercept and intercept-and-trend models to verify that none of the variables were integrated to order two. As observed, all variables become stationary at the first order of integration, with some being stationary at level. This confirms the suitability of the ARDL approach for model cointegration. Table 2: Unit root tests results. Augmented Dickey-Fuller (ADF) Intercept Intercept and trend Variables Level First difference Level First difference GDP −1.069 −12.529∗∗∗ −3.553∗∗ −7.616∗∗∗ PETRC −1.466 −7.675∗∗∗ −3.553∗∗ −7.616∗∗∗ ELEC −0.608 −9.940∗∗∗ −6.004∗∗∗ −9.854∗∗∗ ENRMG −2.785∗ −13.960∗∗∗ −5.653∗∗∗ −13.844∗∗∗ CPI −1.585 −5.770∗∗∗ −3.205∗ −5.983∗∗∗ Phillips-Perron (PP) Intercept Intercept and trend Variables Level First difference Level First difference GDP −1.28 −16.574∗∗∗ −5.402∗∗∗ −16.437∗∗∗ PETRC −1.250 −9.915∗∗∗ −3.627∗∗ −9.508∗∗∗ ELEC −1.208 −21.991∗∗∗ −6.130∗∗∗ −21.845∗∗∗ ENRMG −4.990∗∗∗ −18.313∗∗∗ −5.929∗∗∗ −18.101∗∗∗ CPI −1.543 −6.028∗∗∗ −2.107 −6.134∗∗∗ Kwiatkowski-Phillips-Schmidt-Shin (KPSS) Intercept Intercept and trend Variables Level First difference Level First difference GDP 0.993∗∗∗ 0.070 0.104 0.060 PETRC 0.977∗∗∗ 0.133 0.079 0.100 ELEC 1.000∗∗∗ 0.080 0.138∗ 0.078 ENRMG 0.604∗∗∗ 0.082 0.088 0.077 CPI 1.022∗∗∗ 0.257 0.181∗∗ 0.083 Note: ∗∗∗, ∗∗ et ∗ denote statistical significance at 1%, 5% and 10%, respectively. The lag lengths for the ADF and PP tests are automatically chosen by Schwarz’s Bayesian information criteria. The bandwidth for the KPSS test is automatically selected through the bandwidth selection procedure proposed by Newey & West (1994). The null hypothesis of the ADF and the PP tests is that the series are not stationary while the null hypothesis of the KPSS test is that the series are stationary. 7 4.2 ARDL bounds test The results of the ARDL bounds test are presented in Table 3. It shows that the optimal model de- termined by the Hannan–Quinn information criterion is ARDL(1,2,4,3,0). The test findings reveal that the computed F-test statistic is 9.587319, which is significant and greater than the upper bound value for 5% and 1% statistical significance as specified by Narayan (2005). Therefore, the Bounds test outcome confirms the existence of a cointegration connection among dependent and independent variables at the 1% significance level. Table 3: Bounds F-test outcome. Estimated Model Lag order F-statistics FGDP(GDPt | ELECt,PETRCt,ENRMGt,CPIt) 1,2,4,3,0 9.587319∗∗∗ Critical values for F-statistics Lower bound I(0) Upper bound I(1) 10% 2.323 3.273 5% 2.743 3.792 1% 3.710 4.965 Note: ∗∗∗ denotes statistical significance at 1%. Critical values are taken from Narayan (2005) for a sample size of 60. 4.3 Diagnostic and stability tests Table 4 shows that the ARDL model in this study has met the requirements of standard assumptions because the probability associated with each diagnostic test is greater than 0.1. The diagnostic tests suggest that the residuals are normally distributed, free from serial correlation, and homoskedastic. These results are based on the Jarque-Bera normality test, the Breusch-Godfrey serial correlation LM test, the Breusch-Pagan-Godfrey heteroskedasticity test, and the ARCH test, respectively. Moreover, the Ramsey RESET test confirms that the ARDL model is well-specified. Table 4: Diagnostic tests. Normality Serial correlation Heteroskedasticity Functional form Jarque-Bera LM test Breusch-Pagan-Godfrey ARCH Ramsey RESET χ2 0.345 0.109 9.296 0.589 1.116 Prob. 0.841 0.947 0.811 0.443 0.297 Besides, Figures 1a and 1b depict the results of CUSUM and CUSUMSQ tests. The plotting of CUSUM and CUSUMSQ statistics are within the critical bounds of 5% significant level, which is the significant value line and determines the stability of the estimated methodology. These results mean that the model is valid and robust, and can be used for policy recommendations. 4.4 Coefficient estimation results Table 5, Panel A, presents the long-run coefficients of economic growth with respect to the indepen- dent variables. Notably, both electricity and petroleum consumption positively influence economic growth. On the one hand, a 1% increase in electricity consumption leads to an increase of 0.352% in GDP. Therefore, recognizing the vital role of electricity access in Madagascar’s expanding econ- omy, the government’s focus on promoting a diversified electricity generation mix, prioritizing solar and hydropower, holds the potential to significantly accelerate economic growth if rigorously imple- mented (MEH, 2015; Praene et al., 2017; Rafitoson, 2017). Furthermore, a 1% increase in petroleum 8 Figure 1: CUSUM and CUSUMSQ plots. 20 12 20 13 20 14 20 15 20 16 20 17 20 18 20 19 20 20 20 21 20 22 −20 −15 −10 −5 0 5 10 15 20 CUSUM 5% significance (a) Cumulative sum of recursive residuals. 20 12 20 13 20 14 20 15 20 16 20 17 20 18 20 19 20 20 20 21 20 22 −0.2 0 0.2 0.4 0.6 0.8 1 1.2 CUSUMSQ 5% significance (b) Cumulative sum of squares of recursive residuals. consumption results in a 0.363% increase in GDP. Specifically, prioritizing infrastructures that op- timize petroleum consumption like efficient road networks, strategically located logistics hubs, and modernized processing facilities in mining and other key industries, could drive substantial economic growth across Madagascar. However, neglecting the counterpoints would be shortsighted. Long-term reliance on fossil fuels exposes Madagascar to price volatility, supply chain vulnerabilities, environ- mental degradation, and resource depletion, potentially undermining future economic gains. Moreover, we see that Madagascar’s dependence on imported energy weighs heavily on its GDP. Table 5 reveals a statistically significant negative impact of energy imports on the country’s long-run economic performance. A 1% increase in energy imports, in fact, reduces GDP by 0.301%. This aligns with Murshed et al. (2020)’s findings for Sri Lanka, which attributed the negative impact to the coun- try’s trade balance deficits. Similarly, trade balance deficits have been a persistent challenge for Mada- gascar’s economy throughout its history (Douane Malagasy, 2023; INSTAT, 2022b; INSTAT-MADIO, 1997). Additionally, the observed deficits of trade balance for energy underscore the heavy dependence of its economy on imported energy (Subtil, 2021). To counter this vulnerability and improve economic performance, reducing the demand for imported fuels and diversifying the national energy mix with domestic renewable sources become crucial (Surroop & Raghoo, 2018). Besides, a 1% increase in global prices significantly reduces Madagascar’s GDP growth by 0.143 percentage points, likely due to the country’s heavy reliance on energy imports. This sensitivity un- derscores the trade-off policymakers face: prioritize immediate energy security through continued fossil fuel reliance, or invest in long-term energy diversification towards renewable sources. Balanc- ing energy needs with long-term sustainability considerations will be crucial to ensure Madagascar’s economic trajectory remains resilient to volatile energy markets (Ahuja & Tatsutani, 2009). Table 5 also reports the results of short-run estimates (see Panel B). The results show that the sign of the lagged error-correction term (ECTt−1) is statistically significant at the 1% level with negative sign. The value of the coefficient is −0.975, which suggests that about 97.5% of the last quarter’s disequilibrium is corrected in the current quarter. Therefore, any short-run deviation will take about 1.02 quarters to bring the shock back to equilibrium. Regarding the short-run model’s goodness of fit, the regression results imply that economic growth is moderately explained by the explanatory variables incorporated in the model. The adjusted R-squared reveals that 66.6% of the short-run 9 Table 5: ARDL Long- and short-run results. Panel A. Long-run coefficients Panel B. Short-run coefficients (cointegrating form) Variables Coefficient (std. Error) Variables Coefficient (std. Error) ELECt 0.352 (0.110)∗∗∗ ∆ELECt 0.080 (0.057) PETRCt 0.363 (0.104)∗∗∗ ∆ELECt−1 −0.177 (0.057)∗∗∗ ENRMGt −0.301 (0.117)∗∗ ∆PETRCt 0.328 (0.065)∗∗∗ CPIt −0.143 (0.074)∗ ∆PETRCt−1 −0.131 (0.067)∗ Constant 5.251 (0.350)∗∗∗ ∆PETRCt−2 0.122 (0.070)∗ ∆PETRCt−3 0.166 (0.069)∗∗ ∆ENRMGt 0.135 (0.072)∗ ∆ENRMGt−1 0.350 (0.095)∗∗∗ ∆ENRMGt−2 0.131 (0.076)∗ ECTt−1 −0.975 (0.122)∗∗∗ R-squared 0.717 Adjusted R-squared 0.666 Durbin-Watson 2.011 Note: ∗∗∗, ∗∗ et ∗ denote statistical significance at 1%, 5% and 10%, respectively. variation in real gross domestic product is explained by the explanatory variable. The Durbin-Watson value of 2.011 indicates that the short-run model is free from serial correlation. We note that all short-run coefficients are statistically significant, except for the ∆ELECt coeffi- cient. Electricity consumption at lag (−1) has a significant impact on GDP at the 1% level. The results show that a 1% increase in electricity consumption from the previous quarter leads to a 0.177% de- crease in current GDP. This suggests that increased electricity demand in the previous quarter may be a potential indicator of a decreased efficiency in the production process. If businesses and households aren’t generating sufficient output or value for the electricity consumed, this inefficiency can hinder economic growth. This inefficiency could be attributed to several factors, including outdated and inad- equate electrical infrastructure, distribution loss (Kappen, 2018), and most significantly, fuel supply shortage which is one of the main reasons for the load shedding in Madagascar (Ralitera, 2016, 2023). Load shedding disturbs production processes and disrupts household daily routines. For businesses, this may lead to higher electricity consumption in other part of the day, as they reschedule operations and scramble to catch up on lost production. Households follow a similar pattern, adjusting their rou- tines to compensate for the power cuts. This situation ultimately pushes electricity demand beyond its usual peak periods, leading to higher energy bills for both businesses and households. Our findings for petroleum consumption also show similarities to those for electricity consumption. Petroleum consumption has a significant impact on GDP. The coefficients on the contemporaneous and lagged variables are positive and statistically significant, except for lag (−1) which is negative. This suggests that sudden spikes in petroleum consumption, potentially driven by the need for electricity generation, have short-term negative impacts on economic growth, mirroring the pattern observed with electricity consumption. Furthermore, Madagascar’s transportation sector inefficiencies could partially explain the observed pattern. Outdated and poorly maintained infrastructure, congested roads, and inadequate public transportation contribute to increased fuel consumption (Ralitera, 2016). This, in turn, can disrupt production processes and cause a temporary dip in output. Additionally, our analysis reveals a statistically significant positive short-run relationship between GDP and energy imports. The results show coefficients with positive signs for the contemporaneous and lagged variables. This indicates that increasing energy imports have a direct and positive impact on economic growth in the short run. This finding suggests that increased energy imports, particularly 10 within the near future, can contribute to GDP growth. Again, policymakers must consider the long- term sustainability of this dependence. 4.5 Granger Causality test The results of the VECM Granger test are illustrated in Table 6. Considering the short-run effects, electricity consumption, energy imports, and global prices Granger-cause GDP in the short run, as evidenced by their significant coefficients at the 1% and 10% levels in the GDP equation. This suggests that these variables have a direct and immediate impact on economic growth. Petroleum consumption does not Granger-cause GDP in the short run, as its coefficient is insignificant in the GDP equation. However, petroleum consumption Granger-causes electricity consumption at the 5% level, indicating a unidirectional causal relationship where petroleum consumption influences electricity consumption. Global prices Granger-causes energy imports at the 1% level, suggesting that changes in global energy prices directly affect energy import decisions in the short run. Regarding the long-run results, in the GDP equation, the coefficient on the lagged error correction term is significant with the correct sign at the 1% level, confirming the results from the bounds test for cointegration. This implies long-run Granger causality running from energy demand (including electricity and petroleum consumption and energy imports) and global prices to GDP. This means that these variables have a long-term, interactive influence on economic growth, with adjustments occurring over time to maintain equilibrium. Table 6: Short- and long-run causality tests. Dependent Short-run Long-run variables ∆GDPt ∆ELECt ∆PETRCt ∆ENRMGt ∆CPIt ECTt−1 ∆GDPt − 10.910∗∗∗ 1.018 3.558∗ 3.495∗ −0.772∗∗∗ ∆ELECt 0.630 − 3.897∗∗ 0.117 0.198 0.457 ∆PETRCt 1.017 0.130 − 0.051 0.008 0.172 ∆ENRMGt 0.047 0.109 0.584 − 39.630∗∗∗ 0.035 ∆CPIt 0.142 0.613 0.001 0.030 − 0.012 Note: ∗∗∗, ∗∗ et ∗ denote statistical significance at 1%, 5% and 10%, respectively. 5 Concluding remarks We conducted a new assessment of the impact of energy demand on economic growth in Madagas- car based on 2007Q1–2022Q4 quarterly data from relevant Malagasy authorities. While research on energy and growth specifically in Madagascar remains limited, our study stands out by utilizing exclu- sively quarterly data from Malagasy authorities. Not only does this approach contribute to valuable locally-driven insights, but also bypasses potential biases and discrepancies inherent in using exter- nal data. Moreover, the higher frequency of data enables robust application of econometric techniques, allowing for a more in-depth analysis of the nuanced patterns and relationships between energy and growth in Madagascar. In this study, we applied the ARDL bounds testing approach and found evidence of cointegration among the series. Our results indicate that, in the long run, electricity consumption and petroleum consumption exert positive and significant effects on economic growth, while energy imports and global prices exhibit negative and significant effects. Next, we employed Granger-causality test based on Er- ror Correction Models to examine the causal relationships between energy demand, global prices, and economic growth. The results revealed the existence of a long-run causality running from both energy demand and global prices to economic growth. Additionally, in the short run, electricity consumption, 11 energy imports, and global prices Granger-cause economic growth; petroleum consumption Granger- causes electricity consumption; and global prices Granger-cause energy imports. Our findings confirm that Madagascar is an energy-dependent economy and underscores the cru- cial role of reliable and affordable electricity in powering economic growth. Increasing access to clean electricity can unlock this potential and foster productivity across sectors. However, Madagascar’s dependence on volatile external resources like imported petroleum exposes the country to economic vulnerabilities and hinders sustainable development. Diversifying the energy mix with renewables presents a solution, lessening reliance on imported fuels and mitigating risks associated with price fluctuations and supply disruptions. Fortunately, Madagascar boasts significant potential for renew- able energy generation, particularly solar, wind, and hydropower. Utilizing these resources offers a cost-effective, secure, and environmentally friendly alternative to fossil fuels, paving the way for eco- nomic growth, energy security, and a sustainable future. References AfDB – African Development Bank (2023), “Madagascar Economic Outlook”, Retrieved October 10, 2023, from https://www.afdb.org/en/knowledge/publications/african-economic-outlook Dilip Ahuja & Marika Tatsutani (2009), “Sustainable energy for developing countries”, Surveys and Perspectives Integrating Environment and Society, 2(1), 1–16. Mohammed Alnour (2021), “The dynamic impact of energy consumption on economic growth in Sudan: A vector autoregression analysis”, International Journal of Developing and Emerging Economies, 9(2), 17–47. Fidimanantsoa Andriamanga (2017), “Relation entre l’énergie et la croissance économique : approche empirique appliquée au cas de Madagascar pour la periode 1995 à 2015”, Munich Personal RePEc Archive, No. 83035, University Library of Munich. ARCEB – Projet d’Appui au Renforcement des Capacités d’Analyse des Facteurs de Vulnérabilité Structurelle et la Promotion de l’Économie Bleue (2022), Analyse Approfondie de l’Ensemble des Contraintes qui Freinent le Développement du Secteur de l’Électricité, Ministère de l’Énergie et des Hydrocarbures - Projet ARCEB, Madagascar, 2022. Bebonchu Atems & Chelsea Hotaling (2018), “The effect of renewable and nonrenewable electricity generation on economic growth”, Energy Policy, 112, 111–118. Robert L. Brown, James Durbin & James M. Evans (1975), “Techniques for testing the constancy of re- gression relationships over time”, Journal of the Royal Statistical Society: Series B (Methodological), 37, 149–163. Hyo-Yeon Choi & Seung-Hoon Yoo (2016), “Oil consumption and economic growth: The case of Brazil”, Energy Sources, Part B: Economics, Planning, and Policy, 11(8), 705–710. David A. Dickey & Wayne A. Fuller (1979), “Distribution of the estimators for autoregressive time series with a unit root”, Journal of the American Statistical Association, 74(366), 427–431. Douane Malagasy (2023), “Les statistiques - Commerce extérieur”. Retrieved August 01, 2023, from http://www.douanes.gov.mg/statistiques-et-bilan/ Obas J. Ebohon (1996), “Energy, economic growth and causality in developing countries: A case study of Tanzania and Nigeria”, Energy Policy, 24(5), 447–453. 12 https://www.afdb.org/en/knowledge/publications/african-economic-outlook http://journals.openedition.org/sapiens/823 https://ssrn.com/abstract=3889000 https://ssrn.com/abstract=3889000 https://mpra.ub.uni-muenchen.de/id/eprint/83035 https://mpra.ub.uni-muenchen.de/id/eprint/83035 http://www.mef.gov.mg/assets/vendor/ckeditor/plugins/kcfinder/upload/files/ARCEB/Etude%20Secteur%20Energie%20Rapport%20Final%20_%20Version%20finale%20Novembre%202022.pdf http://www.mef.gov.mg/assets/vendor/ckeditor/plugins/kcfinder/upload/files/ARCEB/Etude%20Secteur%20Energie%20Rapport%20Final%20_%20Version%20finale%20Novembre%202022.pdf https://doi.org/10.1016/j.enpol.2017.10.015 https://doi.org/10.1016/j.enpol.2017.10.015 https://doi.org/10.1111/j.2517-6161.1975.tb01532.x https://doi.org/10.1111/j.2517-6161.1975.tb01532.x https://doi.org/10.1080/15567249.2012.740144 https://doi.org/10.2307/2286348 https://doi.org/10.2307/2286348 http://www.douanes.gov.mg/statistiques-et-bilan/ https://doi.org/10.1016/0301-4215(96)00027-4 https://doi.org/10.1016/0301-4215(96)00027-4 Robert F. Engle and Clive W. J. Granger (1987), “Co-integration and error correction: Representation, estimation, and testing”, Econometrica, 55(2), 251–276. Delphin K. Espoir, Regret Sunge & Frank Bannor (2023), “Economic growth, renewable and nonre- newable electricity consumption: Fresh evidence from a panel sample of African countries”, Energy Nexus, 9, 1–20. Koli Fatai, Les Oxley & Frank G. Scrimgeour (2004), “Modelling the causal relationship between en- ergy consumption and GDP in New Zealand, Australia, India, Indonesia, The Philippines and Thai- land”, Mathematics and Computers in Simulation, 64(3–4), 431–445. Olusegun O. Gbadebo & Chinedu Okonkwo (2009), “Does energy consumption contribute to economic performance? Empirical evidence from Nigeria”, Journal of Economics and International Finance, 1(2), 044–058. Sajal Ghosh (2009), “Import demand of crude oil and economic growth: Evidence from India”, Energy Policy, 37(2), 699-702. Víctor Gómez & Agustín Maravall (1997), Guide For Using The Programs TRAMO and SEATS, Work- ing Papers 9805, Banco de España. George Hondroyiannis, Sarantis Lolos & Evangelia Papapetrou (2002), “Energy consumption and eco- nomic growth: assessing the evidence from Greece”, Energy Economics, 24(4), 319–336. INSTAT – Institut National de la Statistique (2019a), “Comptes nationaux trimestriels | Note méthodologique et résultats”, INSTAT, Direction de la Comptabilité Nationale et de la Modélisa- tion, Madagascar. INSTAT – Institut National de la Statistique (2019b), Note Méthodologique des Séries de Comptes Nationaux Annuels de Madagascar, Base 2007, INSTAT, Direction de la Comptabilité Nationale et de la Modélisation, Madagascar. INSTAT – Institut National de la Statistique (2019c), Le nouvel Indice des Prix à la Consommation, INSTAT, Direction des Statistiques des Conditions de Vie des Ménages, Madagascar. INSTAT – Institut National de la Statistique (2022a), “Indice des prix à la Consommation (IPC)”, Wayback Machine (January 20, 2022). Retrieved August 10, 2023, from https://web.archive.or g/web/20220120074937/https://www.instat.mg/statistiques/bases-de-donnees/nipc INSTAT – Institut National de la Statistique (2022b), “Tableau de Bord de l’Économie”, Wayback Ma- chine (June 02, 2023) Retrieved November 12, 2023, from https://web.archive.org/web/202306 02151447/https://www.instat.mg/thematique/economie INSTAT – Institut National de la Statistique (2023), “Comptes Nationaux”, Wayback Machine (June 02, 2023) Retrieved August 02, 2023, from https://web.archive.org/web/20230602143320/htt ps://www.instat.mg/thematique/comptes-nationaux INSTAT-MADIO – Institut National de la Statistique avec l’appui du projet MADIO (1997), “Evolution trimestrielle du commerce extérieur 1996, 1997”, Projet MADIO, Etude no. 9701E, 9702E, 9703E, 9712E, 9739E, 9740E. JIRAMA – Jiro sy Rano Malagasy (2023), “Consommation d’électricité par branches d’activités (2007– 2022)” [unpublished raw data], JIRAMA, Madagascar. 13 https://doi.org/10.2307/1913236 https://doi.org/10.2307/1913236 https://doi.org/10.1016/j.nexus.2022.100165 https://doi.org/10.1016/j.nexus.2022.100165 https://doi.org/10.1016/S0378-4754(03)00109-5 https://doi.org/10.1016/S0378-4754(03)00109-5 https://doi.org/10.1016/S0378-4754(03)00109-5 https://academicjournals.org/journal/JEIF/article-abstract/59ED4EC1892 https://academicjournals.org/journal/JEIF/article-abstract/59ED4EC1892 https://doi.org/10.1016/j.enpol.2008.10.021 https://www.bde.es/f/webbde/SES/Secciones/Publicaciones/PublicacionesSeriadas/DocumentosTrabajo/98/Fic/dt9805e.pdf https://doi.org/10.1016/S0140-9883(02)00006-3 https://doi.org/10.1016/S0140-9883(02)00006-3 https://web.archive.org/web/20211129105234/https://www.instat.mg/p/comptes-nationaux-trimestriels-note-methodologique-et-resultats https://web.archive.org/web/20211129105234/https://www.instat.mg/p/comptes-nationaux-trimestriels-note-methodologique-et-resultats https://web.archive.org/web/20220120074937/https://www.instat.mg/statistiques/bases-de-donnees/nipc https://web.archive.org/web/20220120074937/https://www.instat.mg/statistiques/bases-de-donnees/nipc https://web.archive.org/web/20230602151447/https://www.instat.mg/thematique/economie https://web.archive.org/web/20230602151447/https://www.instat.mg/thematique/economie https://web.archive.org/web/20230602143320/https://www.instat.mg/thematique/comptes-nationaux https://web.archive.org/web/20230602143320/https://www.instat.mg/thematique/comptes-nationaux https://web.archive.org/web/20111113144818/http://www.instat.mg/pdf/9701e.pdf https://web.archive.org/web/20111113144818/http://www.instat.mg/pdf/9702e.pdf https://web.archive.org/web/20111113144818/http://www.instat.mg/pdf/9703e.pdf https://web.archive.org/web/20111113144818/http://www.instat.mg/pdf/9712e.pdf https://web.archive.org/web/20111113144818/http://www.instat.mg/pdf/9739e.pdf https://web.archive.org/web/20111113144818/http://www.instat.mg/pdf/9740e.pdf Mulugeta S. Kahsai, Chali Nondo, Peter V. Schaeffer & Tesfa G. Gebremedhin (2012), “Income level and the energy consumption–GDP nexus: Evidence from Sub-Saharan Africa”, Energy Economics, 34(3), 739–746. Jan F. Kappen (2018), Madagascar – Least-Cost Electricity Access Development Project (English), Washington, D.C.: World Bank Group. Fatih Karanfil & Yuanjing Li (2015), “Electricity consumption and economic growth: Exploring panel- specific differences”, Energy Policy, 82, 264–277. Paul A. Kwakwa (2012), “Disaggregated energy consumption and economic growth in Ghana”, Inter- national Journal of Energy Economics and Policy, 2, 34–40. Denis Kwiatkowski, Peter C. B. Phillips, Peter Schmidt & Yongcheol Shin (1992), “Testing the null hypothesis of stationarity against the alternative of a unit root: How sure are we that economic time series have a unit root?”, Journal of Econometrics, 54(1–3), 159–178. MEH – Ministère de l’Énergie et des Hydrocarbures (2015), Lettre de Politique de l’Énergie de Mada- gascar 2015–2030, MEH, Madagascar. Nanzia F. Mmbaga, Yusuph Kulindwa & Isaac Kazungu (2023), “Energy consumption and economic growth nexus in East African sub-region: Interactive dynamics of human capital”, African Journal of Economic Review, 11(5), 50–69. Muntasir Murshed, Haider Mahmood, Tarek T. Y. Alkhateeb & Mohga Bassim (2020), “The impacts of energy consumption, energy prices and energy import-dependency on gross and sectoral value-added in Sri Lanka, Energies, 13(24), 1–22. Kameshnee Naidoo & Christiaan Loots (2020), Madagascar. Energy and the Poor: Unpacking the In- vestment Case for Clean Energy, UN Capital Development Fund (UNCDF). Paresh K. Narayan (2005), “The saving and investment nexus for China: evidence from cointegration tests”, Applied Economics, 37(17), 1979–1990. Paresh K. Narayan & Baljeet Singh (2007), “The electricity consumption and GDP nexus for the Fiji Islands”, Energy Economics, 29(6), 1141–1150. Whitney K. Newey & Kenneth D. West (1994), “Automatic lag selection in covariance matrix estima- tion”, The Review of Economic Studies, 61(4), 631–653. Bui H. Ngoc (2019), “Energy consumption and economic growth nexus in Vietnam: An ARDL ap- proach”, In: Vladik Kreinovich, Nguyen N. Thach, Nguyen D. Trung, Dang V. Thanh (Eds), Be- yond Traditional Probabilistic Methods in Economics, Studies in Computational Intelligence, 809, pp. 311–322, Springer. Chali Nondo, Mulugeta S. Kahsaib & Peter V. Schaeffer (2010), “Energy Consumption and Economic Growth: Evidence from COMESA Countries”, Regional Research Institute Working Papers, 2010-1, West Virginia University. Nicholas M. Odhiambo (2009), “Energy consumption and economic growth nexus in Tanzania: An ARDL bounds testing approach”, Energy Policy, 37(2), 617–622. Mawussé K. N. Okey (2009), “Consommation d’énergies et croissance du PIB dans les pays de l’UEMOA: Une analyse en données de panel”, Munich Personal RePEc Archive, No. 15521, Uni- versity Library of Munich. 14 https://doi.org/10.1016/j.eneco.2011.06.006 https://doi.org/10.1016/j.eneco.2011.06.006 http://documents.worldbank.org/curated/en/300031518715365390/Madagascar-Least-Cost-Electricity-Access-Development-Project https://doi.org/10.1016/j.enpol.2014.12.001 https://doi.org/10.1016/j.enpol.2014.12.001 https://dergipark.org.tr/en/pub/ijeeep/issue/31899/350662 https://doi.org/10.1016/0304-4076(92)90104-Y https://doi.org/10.1016/0304-4076(92)90104-Y https://doi.org/10.1016/0304-4076(92)90104-Y http://www.ore.mg/Publication/Rapports/LettreDePolitique.pdf http://www.ore.mg/Publication/Rapports/LettreDePolitique.pdf https://www.ajol.info/index.php/ajer/article/view/260806 https://www.ajol.info/index.php/ajer/article/view/260806 https://doi.org/10.3390/en13246565 https://doi.org/10.3390/en13246565 https://doi.org/10.3390/en13246565 https://www.uncdf.org/article/6474/energy-and-the-poor-unpacking-the-investment-case-for-clean-energy https://www.uncdf.org/article/6474/energy-and-the-poor-unpacking-the-investment-case-for-clean-energy http://dx.doi.org/10.1080/00036840500278103 http://dx.doi.org/10.1080/00036840500278103 https://doi.org/10.1016/j.eneco.2006.05.018 https://doi.org/10.1016/j.eneco.2006.05.018 https://doi.org/10.2307/2297912 https://doi.org/10.2307/2297912 https://doi.org/10.1007/978-3-030-04200-4_24 https://doi.org/10.1007/978-3-030-04200-4_24 https://researchrepository.wvu.edu/rri_pubs/87 https://researchrepository.wvu.edu/rri_pubs/87 https://doi.org/10.1016/j.enpol.2008.09.077 https://doi.org/10.1016/j.enpol.2008.09.077 https://mpra.ub.uni-muenchen.de/15521/ https://mpra.ub.uni-muenchen.de/15521/ OMH – Office Malgache des Hydrocarbures (2023a), “Bulletin Pétrolier”. Retrieved August 10, 2023, from http://www.omh.mg/index.php?idm=5&CL=bulpetro OMH – Office Malgache des Hydrocarbures (2023b), “Ventes nationales de produits pétroliers (2007– 2022)” [Unpublished raw data], OMH, Madagascar. OMH – Office Malgache des Hydrocarbures (2023c), “Ventes nationales de produits pétroliers par composante (2016–2022)” [Unpublished raw data], OMH, Madagascar. Adegbemi B. Onakoya, Adegbemi O. Onakoya, Olalekan A. Jimi-Salami & Babatunde O. Odedairo (2013), “Energy consumption and Nigerian economic growth: An empirical analysis”, European Sci- entific Journal, 9(4), 25–40. Hashem M. Pesaran & Yongcheol Shin (1999), “An autoregressive distributed-lag modelling approach to cointegration analysis”, In: Steinar Strøm (Ed.), Econometrics and Economic Theory in the 20th Century: The Ragnar Frisch Centennial Symposium, pp. 371–413, Cambridge University Press. Hashem M. Pesaran, Yongcheol Shin & Richard J. Smith (2001), “Bounds testing approaches to the analysis of level relationships”, Journal of Applied Econometrics, 16, 289–326. Peter C. B. Phillips & Pierre Perron (1988), “Testing for a unit root in time series regression”, Biometrika, 75(2), 335–346. Jean Philippe Praene, Mamy H. Radanielina, Vanessa R. Rakotoson, Ando L. Andriamamonjy, Frantz Sinama, Dominique Morau & Hery T. Rakotondramiarana (2017), “Electricity generation from re- newables in Madagascar: Opportunities and projections”, Renewable and Sustainable Energy Re- views, 76, 1066–1079. Ketakandriana Rafitoson (2017), La lente marche vers la transition énergétique à Madagascar: État des lieux et perspectives, Friedrich-Ebert-Stiftung, Madagascar. Miangaly Ralitera (2016), “Le délestage frappe faute de carburants”, L’Express de Madagascar. Re- trieved December 10, 2023, from https://lexpress.mg/13/10/2016/le-delestage-frappe-fau te-de-carburants/ Miangaly Ralitera (2022), “Infrastructures – Les routes nationales impraticables”, L’Express de Mada- gascar. Retrieved December 10, 2023, from https://lexpress.mg/11/01/2022/infrastructures -les-routes-nationales-impraticables/ Miangaly Ralitera (2023), “Approvisionnement en éléctricité – Délestage tournant, faute de gasoil”, L’Express de Madagascar. Retrieved December 10, 2023, from https://lexpress.mg/15/04/2023 /approvisionnement-en-electricite-delestage-tournant-faute-de-gasoil/ Molem C. Sama & Ndifor R. Tah (2016), “The effect of energy consumption on economic growth in Cameroon”, Asian Economic and Financial Review, 6(9), 510–521. Colin Subtil (2021), “Madagascar: le retour de la stabilité politique permettra-t-il le décollage de l’économie malgache ?”, MacroDev, 33, Éditions AFD, 1–44. Dinesh Surroop & Pravesh Raghoo (2018), “Renewable energy to improve energy situation in African island states”, Renewable and Sustainable Energy Reviews, 88, 176–183. Amélie Voninirina & Saminirina Andriambelosoa (2014), “Étude sur l’énergie à Madagascar”, Cahier de Recherche en Analyse Économique, 21, Centre de Recherches, d’Études et d’Appui à l’Analyse Économique de Madagascar (CREAM), Madagascar. 15 http://www.omh.mg/index.php?idm=5&CL=bulpetro https://eujournal.org/index.php/esj/article/view/765 https://doi.org/10.1017/CCOL521633230.011 https://doi.org/10.1017/CCOL521633230.011 https://doi.org/10.1002/jae.616 https://doi.org/10.1002/jae.616 https://doi.org/10.1093/biomet/75.2.335 https://doi.org/10.1016/j.rser.2017.03.125 https://doi.org/10.1016/j.rser.2017.03.125 https://library.fes.de/pdf-files/bueros/madagaskar/15155.pdf https://library.fes.de/pdf-files/bueros/madagaskar/15155.pdf https://lexpress.mg/13/10/2016/le-delestage-frappe-faute-de-carburants/ https://lexpress.mg/13/10/2016/le-delestage-frappe-faute-de-carburants/ https://lexpress.mg/11/01/2022/infrastructures-les-routes-nationales-impraticables/ https://lexpress.mg/11/01/2022/infrastructures-les-routes-nationales-impraticables/ https://lexpress.mg/15/04/2023/approvisionnement-en-electricite-delestage-tournant-faute-de-gasoil/ https://lexpress.mg/15/04/2023/approvisionnement-en-electricite-delestage-tournant-faute-de-gasoil/ https://doi.org/10.18488/journal.aefr/2016.6.9/102.9.510.521 https://doi.org/10.18488/journal.aefr/2016.6.9/102.9.510.521 https://doi.org/10.3917/afd.subtil.2021.01.0001 https://doi.org/10.3917/afd.subtil.2021.01.0001 https://doi.org/10.1016/j.rser.2018.02.024 https://doi.org/10.1016/j.rser.2018.02.024 https://web.archive.org/web/20150404022515/http://www.cream.mg/pub/cahier21.pdf Introduction Brief empirical review Data and methodology Data descriptions Empirical methodology Empirical results Unit root tests ARDL bounds test Diagnostic and stability tests Coefficient estimation results Granger Causality test Concluding remarks