Energy-GDP Nexus for Oil-Exporting Country: The Case of Bahrain

The main objective of this paper is to investigate the relationships between electricity consumption and economic growth in Bahrain between 1980 and 2019, utilizing the Autoregressive Distributed Lag (ARDL) model. Estimates revealed the existence of a long run equilibrium relationship between the electricity consumptions per capita and GDP per capita. The Pairwise Granger Causality Test results indicated a unidirectional causality running from electricity consumption to economic growth. The estimation results provide a strong support for the Growth hypothesis in Bahrain, suggesting that Bahrain economy is Energy-dependent. Therefore, it should be taken into consideration that energy consumption has effect on economic growth rates by policy makers.


INTRODUCTION
Since energy is a key source in production and in many consumption activities, the causal relationship between energy consumption and economic growth has been the subject of intense research and empirical studies over different countries and regions around the world, using different econometric techniques (e.g. ECM, ARDL, VAR, OLS-EG, DOLS, FMOLS, etc.). On the energy economics, four hypotheses have been widely subject to testing. Sometimes these hypotheses are named neutrality, conservation, growth and feedback hypotheses. As shown soon in the literature, each one of these hypothesis has different implications for designing suitable energy policy for each given country.
In fact, the pioneer empirical work by Kraft and Kraft (1978), in which causality was found to run from GNP to energy consumption in USA. Yet, the question, whether or not energy consumption promotes economic activates or economic growth is still a controversial issue among applied and theoretical researchers. Recent studies on the relationship between the two variables have provided mixed results up till date. However, according to Abosedra et al. (2009) and Chen et al. (2007) this inconsistency is due to differences in data set, different analytical and econometric methods utilized, country and reigns (developed or developing country) and model variables used. The findings from the previous studies vary not only across countries, but depend also on methodologies within the same country (Soytas and Sari, 2003). While a country-specific causality study between energy consumption and economic growth can provide insight for the design of future energy policies, it is also important to reach unambiguous results for policy implementation (Erol and Yu 1987). Therefore, it is important for policy makers to have a clear understanding of the causal relationship between energy consumption and economic growth.
in the country's GDP. The non-oil GDP accounted for about 80 percent of total GDP. It dominated by infrastructure and services sectors, has recorded tremendous growth over time (Kreishan et al., 2018). It has also been argued that the possible impact of energy consumption on growth will depend on the structure of the economy and the level of economic growth of the country concerned (Naser, 2017). As the economy grows, its production structure is likely to shift towards services, which are not energy intensive activities (Solow, 1978, Denison, 1985, Cheng, 1995and Belloumi, 2009). Bahrain has become the center of finance in the region, thus is an essential factor for the effectiveness of the banking and other services sectors. The role of electricity for Bahrain economy seems to be crucial to continue the country plan of diversification. In fact without electricity Bahrain will not be able to achieve this goal. Currently, per capita energy consumption among Bahrainis is one of the highest in the world, it uses two folds more energy per capita than Japan. In 2018, an average Bahraini consumed 9456 kg oil equivalent of energy. By comparison, Japan and the US have per capita energy uses of 4033 and 7747 kg, respectively (World Development Indicator, 2019).
This study attempts to examine empirically the existence and direction of causal relationship between electricity consumption and economic growth rate as proxy for economic activates in Bahrain for the period between 1980 and 2019. Doing this study for Bahrain can be justified on three reasons: examining the role of energy consumption in stimulating economic growth in Bahrain, testing the direction of causality between energy consumption and economic growth and finally investigating which one of the four above-mentioned hypotheses is applicable for Bahrain. Therefore, the empirical results of this study might provide policymakers in Bahrain with some insights towards formulating appropriate energy development policies in Bahrain. Finally, the study uses the ARDL methodology of cointegration, as recommended by Ozturk (2010) and (Emeka and Aham, 2016) which has a number of advantages over other methods of cointegration.
The paper is organized as follows: Section 2 reviews the literature on the subject. A discussion on the methodology and description of the data used are presented in Section 3. Section 4 reports on empirical results. Conclusions of the study and the policy implications are produced in Section 5.

LITERATURE REVIEW
The debate among energy economists about the relationship between energy consumption and GDP has led to emergence of four hypotheses which has been widely subject to testing. As summarized in Payne (2009) the first one is the Neutrality hypothesis which suggests that energy consumption has only minor effect on the production of real output. It declares that there is no causality in either direction between economic growth and electricity consumption. Therefore, neither conservative nor expansive energy consumption policies have any effect on economic growth. The second hypothesis which might be called Conservation hypothesis suggests that the economy is less dependent on energy (Hwang and Gum 1992). Consequently, implementing an energy conserving policy does not have an adverse impact on economic growth. Therefore it presumes the existence of uni-directional causality running from economic growth to energy consumption. The Growth hypothesis however, suggests that energy consumption affects economic growth directly, implying that there is causality from energy consumption to economic activity. Accordingly, implementing an energy conserving policy will slow economic growth. This hypothesis, posits that unidirectional causality running from energy consumption to economic growth. It can be theoretically inferred that the economy is energy-dependent. The fourth mainstream known as the Feedback hypothesis posits that a bi-directional causality between energy consumption and economic growth, which implies that energy consumption and economic growth complement each other or are determined together.
Empirically, various studies have focused on different countries, time periods and have used different proxy variables for both energy consumption and economic activates. In general, the empirical work could be classified into two groups: a countryspecific studies and multi-country studies. Among multi-country studies, Hossein et al. (2012) test the Granger causality between energy consumption and economic growth for Organization of Petroleum Exporting Countries (OPEC) countries for the period 1980-2008. They conclude that reducing energy consumption will not negatively affect economic growth rather it will reduce CO 2 emissions. Narayan and Popp (2012) investigate the long-run impact of energy consumption on real GDP for 93 countries over the period 1980-2006 using panel fully modified ordinary least squares. They find that the impact is minimal for all the panels, therefore energy conservation policies will benefit some but not all the countries.

Asafu-Adjaye (2000) used cointegration and Error Correction
Modeling techniques to estimate the causal relationships for 6 Asian countries. His results indicated a short-run unidirectional Granger causality running from energy to GDP for India and Indonesia, but a bidirectional relationship for the other four countries. Moreover, the author includes energy prices in his study, and finds that energy, income, and prices are mutually causal for the Philippines and Thailand. For India and Indonesia, however, the causality is unidirectional, running from energy and prices to income. However, Asafu-Adjaye's empirical results for Indonesia and the Philippines are different from other studies (Yu and Choi, 1985;Cheng and Lai, 1997;Masih and Masih, 1998 ;Chen et al., 2007;James, 2009) using single data set and panel data procedure for 10 industrializing and developing Asian countries, showed a unidirectional short-run causality running from economic growth to electricity consumption in the panel data. Ozturk and Acaravci (2011) investigate the short-run and long-run causality between electricity consumption and economic growth for 11 selected Middle East and North Africa (MENA) countries using (ARDL) bounds testing approach of cointegration and vector errorcorrection models. The overall results of this study indicate that there is no relationship between the electricity consumption and the economic growth in most of the MENA countries.
At the level of single countries, Ghosh (2002) has applied Granger causality test on the bivariate vector autoregressive model (VAR) to test for causal relationship for India. He focuses on a particular form of energy, i.e. electricity. The results indicate that there is unidirectional Granger causality running from economic growth to electricity consumption without any feedback effect. However, Ghosh (2009) has re-examined the relationship between electricity supply, employment and real GDP for India within a multivariate framework using (ARDL). Although a long-run equilibrium relationship has been established among these variables, the study found absence of causality running from electricity supply to real GDP. Shiu and Pun-Lee (2004) investigate the causal relationship between electricity consumption and real GDP for China during 1971-2000. The results indicate that real GDP and electricity consumption for China are cointegrated and there is unidirectional Granger causality running from electricity consumption to real GDP but not vice versa. Baranzini et al. (2013) investigate the relationship between energy use and economic growth for Switzerland over the period 1950-2010 using (ARDL). They find that conserving energy policies do not necessarily have negative impact on Swiss economic growth. Shahbaz and Feridun (2012) using (ARDL) approach, found a unidirectional causality running from economic growth to electricity consumption in Pakistan. In Malaysia, Abosedra et al. (2009) utilize the Granger-Causality method within VAR framework, revealed unidirectional causality running from electricity consumption to economic growth in Malaysia. Adebola and Opeyemi (2011) using the (ARDL) approach to cointegration, showed that there is a long-run causal relationship running from electricity to economic growth in Nigeria.
Looking at the Arab region the empirical results are still mixed, for example Belloumi (2009) has studied the causal relationship between energy consumption and GDP for Tunisia during the period 1971-2004. The estimation results indicate that there is a long-run bi-directional causal relationship between the two series and a short-run unidirectional causality from energy to (GDP). Ajlouni (2015) employs ARDL bounds test to examine the shortrun and long-run relationship between energy consumption and economic growth in Jordan using annual data over the period 1980-2012. The result demonstrates a positive bidirectional relationship between energy consumption and economic growth supporting a feedback hypothesis.
On the Gulf countries, Al-Iriani (2006) investigates the causality relationship between (GDP) and energy consumption in the six countries of the Gulf Cooperation Council (GCC). The results indicate a unidirectional causality running from GDP to energy consumption. Evidence shows no support for the hypothesis that energy consumption is the source of GDP growth in the GCC countries. In the same context, Hamdi et al. (2014) explores the relationship between electricity consumption, foreign direct investment, capital and economic growth in the case of Bahrain. Using (ARDL) a cointegration relationship has been detected among the series. It is found that electricity consumption, foreign direct investment and capital add in economic growth.
As highlighted above, the results are varying not only across countries but also on methodologies and time period within the same country. According to Masih and Masih (1997) the factors that produce conflicting findings are methodological differences, definitional specifications of the variables, as well as the type of causality techniques, tests, and lag structures employed in these studies. On the other hand, Ozturk (2010) agrees with the conclusion of Masih and Masih (1997) and adds that the empirical findings are mixed and sometimes contradictory for the same country. He attributes this to differences in data sets, econometric methods and to the variables used in these empirical models, in addition to differences in countries' characteristics. Accordingly, it is worth having an in-depth investigation of the relationship between electricity consumption and economic activities in order to make appropriate energy policies.

METHODOLOGY AND DATA
In order to establish the Energy-GDP nexus in Bahrain economy, the study uses secondary data collected from World Development Indicator's (WDI), in addition to Bahrain Electricity and Water Authority (EWA) and Central Bank of Bahrain (CBB). The dataset starts with 1980 and ends in 2019. For uniformity in measurement and clarity in the interpretation of results; the natural log differences approach has been used to compute and eliminate any serial correlation that might be exist between the variables under study. The following equation has been applied for each variable.
Where (X it) is the value of the variable for current year (t), while (X it−1 ) is the value of the variable for previous year. The study uses electric power consumption per capita denoted by (LECp t ) and measured by KWh per capita, and Real GDP per capita (2005=100) denoted by (LRGP t ) as the proxies for energy consumption and economic growth respectively. Figure 1 illustrates the values of both variables-before taking the logarithm-during the period , where the average electricity consumption per capita during the period of the study is 17673.3 Kwt, with a maximum of 22845.30 Kwt in 2019 and a minimum of 4612.5 Kwt in 1980. The average growth rate electricity consumption per capita is 6.2%. Regarding to the Real GDP per capita, the mean is $14711.90, with a maximum of $24989.40 and a minimum of $7041.600 in 2104 and 1986 respectively. The average Real GDP per capita is 3.04%.
The following two equations are specified for testing the existence and direction of causal relationship between electricity consumption and economic growth: In order to achieve the objective of this study, we have examined the time series properties of the data and establish their order of integration by using Augmented Dickey Fuller (ADF), Phillips and Perron (PP) tests, and Kwiatkowski et al. (KPSS). The study uses the minimum Akaile's Information Criteria (AIC) to determine the appropriate number of lags before performing these tests. The study employs the tests that include intercept and trend term at the level and first differences in case the variables have unit root at level (Im et al., 1995 If the study variables are stationary, the study will apply one of the cointegration tests which is (ARDL) cointegration technique or Bound test of cointegration Shin, 1999) in order to detect the presence of steady state equilibrium between variables by regressing the above equations (2) and (3). After estimating the two equations; the long run relationship of the underlying variables is detected through comparing the F-statistic (Wald test) with critical values (Engle and Granger, 1987). The lower bound critical values assumed that the explanatory variables are integrated at order I(0), while the upper bound critical values assumed that the explanatory variables are integrated at order I(1). When the F-statistic is greater than the upper bound value; this indicates that the Energy-GDP nexus in Bahrain economy has long run relationship, but if the F-statistic falls between the upper and lower bounds values; the results are inconclusive.
The study selected (ARDL) model rather than other tests such as (residual-based Engel and Granger test and the maximum likelihood-based by Johansen) because (ARDL) can be used irrespective whether the variables are Stationary I (0) or integrated of order one I (1) or mixed of both, but it will not apply if they are I (2). This helps to avoid the pretesting problems that related to standard cointegration analysis that needs to classify variables into I (0) and I (1). This means that the procedures of (ARDL) test do not require the pre-testing of the study varibles for unit roots and is robust when there is a single long run relationship between these variables. In addition to when the F-stat. establishes that there is a single long run relationship and the sample data size is small or finite, the error correction representation of (ARDL) becomes relatively more efficient (Emeka and Aham, 2016).
The direction of causality between the two varibles will be examined by Granger causality test (1969) that determines whether 1 time series of a variable is useful to predict another variable. In other word, variable (X) is said to cause variable (Y) if the history of (X) can explain variation in (Y). The test involves estimating the following pair of regressions, assuming that the disturbances (u 1t ) and (u 2t ) are uncorrelated (Gujarati and Porter, 2009).
Since we have two variables, we are dealing with bilateral causality. Equations postulate that current (Y) is related to past values of itself as well as that of (X), and postulates a similar behavior for (X). Fisher statistic is used to recognize the causality direction between both varibles by comparing the p-value with 5% significant level (Rod and Glenn, 1984). To complete the interaction analysis between the two variables, and based on the results of causality test; the study will investigate the impulse response reaction one variable in relation to other variable changes (Koop et al., 1996;Aktan, 2018). Impulse response function represents "a method employed to show the responsiveness of one variable to shocks a function of time or some other independent variables" (Hatemi, 2014, p. 22). The study will trace the effect of one variable on another variable. The main assumption of the impulse analysis is that a shock happens only in one variable at 1 time (David, 2011). Figure 2 Presents the main statistics associated to the study variables, and it shows the variability for both variables. Skewness index and probability of Jarque-Bera normality test show the normality of the variables.   null hypothesis that means underlying series are stationary at first differences, and they are integrated of order one I (1). According to Pesaran et al. (2001) the cointegration test can be carried out because both variables are not I (2) or higher.

Cointegration Test
In order to detect the presence of state equilibrium between the two variables, the study employs (ADRL) bounds test for cointegration by computing bounds F-test in equations (2) and (3). Table 2 shows that the F-statistic exceeds the upper bound value critical value at 5% in both equations, which indicates there is long run relationship between electricity consumption and economic growth. Table 3 illustrates the estimated long run coefficients using ARDL, and it revealed a long run relationship in both equations. The long run coefficient of Real GDP per capita variable in equation (2) is positive and has significant impact on KWh per capita at 5% level. A 1% increase in Real GDP per capita leads to increasing the electricity consumption by 0.52%. Equation (3) shows the importance of electricity to economic growth in Bahrain, where electricity consumption variable is positive and it has significant effect on Real GDP per capita. The increasing in electricity consumption by 1% will increase economic growth by 0.87%.

Causality Tests
To test for the existence and the direction of causality, the study employed Pairwise Granger Causality Tests. Table 4 results suggest rejecting the null hypothesis that states that, electricity consumption does not Granger Cause economic growth, and accepting the alternative hypothesis which indicates there is causality from electricity consumption to economic growth. On the other hand, we accept the null hypothesis that states that, economic growth does not Granger Cause electricity consumption. Accordingly, there is unidirectional causality running from electricity consumption to economic growth. These results consist with the third hypothesis that called "the Growth hypothesis" that suggest that electricity   consumption affects economic growth directly, which inferred that the Bahrain economy is Energy-dependent.

Impulse Response
According to causality results, we can forecast the effect and the impulse response function of electricity consumption on economic growth for the next 10 years. Figure 3 shows the impulse response functions where the adjustment process of economic growth could be completed within these 10 years.

CONCLUSION AND POLICY IMPLICATION
This study investigates the relationship between energy consumption and economic growth rate using Bahrain case during the period from 1980 to 2019. According to the ADF, PP and KPSS unit root test results, we find that both series are integrated of order one I (1). In order to detect the presence of state equilibrium between the two variables, the study employs (ADRL) bounds test for cointegration by computing bounds F-test. The test predicts that such relationship indeed exists. The estimated long run coefficients using (ARDL) revealed that electricity consumption variable is positive and it has significant effect on Real GDP per capita. Thus increasing in electricity consumption by 1% will increase economic growth by 0.87%.
As for causality relationship, we conclude that there is a unidirectional causality relationship between the energy consumption and economic growth in Bahrain during the period from 1980 to 2019. According to results there is causality relationship is from energy consumption to economic growth that means changes in energy consumption affects economic activates. Consequently, it should be taken into consideration that energy consumption has effect on economic growth rate by policy makers in Bahrain.