UNIVERSITY OF CAPE COAST ELECTRICITY CONSUMPTION AND ECONOMIC GROWTH IN GHANA BY ISAAC KWESI AMPAH Thesis submitted to the Department of Economics of the Faculty of Social Sciences, University of Cape Coast, in partial fulfilment of the requirements for award of Master of Philosophy degree in Economics. SEPTEMBER 2012 i    DECLARATION Candidate’s Declaration I hereby declare that this thesis is the result of my own original work and that no part of it has been presented for another degree in this university or elsewhere. Candidate Signature:……………………… Date:……………………. Name: Isaac Kwesi Ampah Supervisors’ Declaration We hereby declare that the preparation and presentation of the thesis were supervised in accordance with the guidelines on supervision of thesis laid down by the University of Cape Coast. Principal Supervisor’s Signature:……………… Date:…………….………….. Name: Dr. Camara Obeng Co-Supervisor’s Signature:……………………. Date:………………………... Name: Dr. Samuel k. Annim ii    ABSTRACT The study examined the relationship between electricity consumption and economic growth in Ghana using annual data for the period 1971 to 2010 by employing autoregressive distributed lag (ARDL) approach to cointegration. The study found a cointegrating relationship among the variables when real GDP was used as the dependent variable and no cointegrating relationship among the variables when electricity consumption was used as the dependent variable. The bounds test results revealed that electricity consumption exerted a positive and statistically significant effect on economic growth both in the short-run and long- run suggesting that higher electricity consumption is crucial to economic growth in Ghana. Further, financial development, labour force and capital stocks exerted a positive and statistically significant impact on economic growth both in the short- run and long-run signifying that capital, labour force and financial development are critical in enhancing sustained economic growth and development in Ghana while inflation revealed a negative effect on economic growth as expected and that it is recommended that government consider creating and maintaining macroeconomic stability necessary for energy development and competition while encouraging the stakeholders in the electricity sector to enhance technological and human resource development.  The Granger causality test result also revealed a unidirectional causality running from economic growth to electricity consumption indicating that electricity conservation policies are viable options for Ghana since this will not retard growth in the economy.  iii    ACKNOWLEDGEMENTS My sincere gratitude goes to my Principal Supervisor, Dr. Camara Obeng and Co-supervisor, Dr. Samuel K. Annim for their constructive criticisms, comments, suggestions, guidance and counselling that helped shape this work. I would also like to express my gratitude to African Economic and Research Consortium (AERC) and the entire staff of Department of Economics University of Cape Coast for their support and motivation throughout the program. Finally, I wish to show my appreciation to my family and to all individuals who contributed in diverse ways toward the successful completion of this thesis. iv    DEDICATION To my sweet mother for her unflinching support, love and inputting into my education. v    TABLE OF CONTENTS Content Page DECLARATION ii ABSTRACT iii ACKNOWLEDGEMENTS iv DEDICATION v TABLE OF CONTENTS vi LIST OF TABLES x LIST OF FIGURES xi LIST OF ACRONYMS xii CHAPTER ONE: INTRODUCTION 1 Background to the Study 1 Statement of the Problem 5 Objectives of the Study 6 Hypotheses 7 Significance of the Study 7 Scope of the Study 8 Organisation of the Study 9 CHAPTER TWO: REVIEW OF RELATED LITERATURE 10 Introduction 10 Overview of the Ghanaian Electricity Sector and Economic growth 10 vi    History of Electrical Power Generation in Ghana 14 Electricity Demand and Supply in Ghana 16 Environment and Energy Policy Issues 19 Theoretical Literature Review 21 The Mainstream Theory of Growth 21 Energy Consumption and Economic Growth 28 Empirical Literature Review 31 Conclusion 59 CHAPTER THREE: METHODOLOGY 60 Introduction 60 Model Specification 60 A prior Expected Signs 61 Justification and Measurement of Variables 62 Data Type and Source 67 Estimation Procedures 68 Data Analysis 78 Conclusion 78 CHAPTER FOUR: RESULTS AND DISCUSSION 79 Introduction 79 Descriptive Statistics 79 Stationarity Test 81 vii    Cointegration Analysis 82 Results of the Long-run Relationship 85 Results of the Short Run Dynamic Model 91 Model Diagnostics and Stability Tests 95 Granger Causality Test 97 Conclusion 100 CHAPTER FIVE: SUMMARY, CONCLUSIONS AND RECOMMENDATIONS 102 Introduction 102 Summary of the Study 102 Conclusions 106 Policy Recommendations 107 Limitations of the Study 109 Future Direction of Research 110 REFERENCES 111 APPENDICES 127 A: Results of plots of variables in levels 127 B: Results of plots of variables in first difference 128 C: Test for the order of integration (ADF): 129 Level and first difference with intercept D: Test for the order of integration (ADF): 130 Level and first difference with intercept and Trend viii    E: Test for the order of integration (Phillips Perron): 131 Level and first difference with intercept F: Test for the order of integration (Phillips Perron): 132 Level and first difference with intercept and Trend G: Plot of cumulative sum of recursive residuals 133 H: Plot of cumulative sum of squares of recursive residuals 134 I: Data used for the Study 135 ix    LIST OF TABLES Table Page 1 Summary Statistics 80 2 Bounds Test for the existence of Cointegration 84 3 Long-run estimates based on SBC-ARDL Approach 86 4 Estimated Short-Run ECM using the ARDL 92 5 Model Diagnostics and Goodness of Fit 95 6 Pairwise Granger Causality Tests 97 x    LIST OF FIGURES Figure Page 1 The Electric Power Flow – from Generation to End Customers 11 xi    LIST OF ACRONYMS ADF Augmented Dickey-Fuller Test AERC African Economic Research Consortium AIC Akaike Information Criterion ARDL Auto Regressive Distributed Lag ASEAN Association of South East Asian Nations BIC Bayesian Information Criterion BW Bandwidth CPI Consumer Price Index EC Electricity Consumption ECG Electricity Company of Ghana ECM Error Correction Model ECT Error Correction Term FD Financial Development GDP Gross Domestic Product GNP Gross National Product GoG Government of Ghana GRIDCo Ghana Grid Company GWh Giga Watts per hour HP Hodrick Prescott IEA International Energy Agency IMF International Monetary Fund xii    INF Inflation ISSER Institute for Statistical, Social and Economic Research JAIDS Journal of Acquired Immune Deficiency Syndrome KWh Kilo Watts per hour MENA Middle East and North Africa countries MW Mega Watts NED Northern Electricity Department OECD Organisation for Economic Co-operation and Development OLS Ordinary Least Square PP Phillips-Perron PPP Public Private Partnership PURC Public Utilities Regulatory Commission  R&D Research and Development RGDP Real Gross Domestic Product RGNP Real Gross National Product SBC Schwartz-Bayesian Criterion SHEP Self Help Electrification Projects SIC Schwarz Information Criterion TAPCo Takoradi Thermal Power Company TFP Total Factor Productivity TICo Takoradi International company TPPS Takoradi Thermal Power Station UCEM Unrestricted Error Correction Model xiii    xiv    USA United States of America VAR Vector Autoregressive VEC Vector Error Correction VECM Vector Error Correction Model VOCs Volatile Organic Compounds VRA Volta River Authority WDI World Development Indicator CHAPTER ONE INTRODUCTION Background to the Study Electricity is a vital ingredient in the socio-economic development of any nation, especially developing nations. Not only is it a critical factor in their development, it is sometimes seen as a measure of the quality and standard of living of a nation. It enables consumers to use daily appliances such as computers, medical devices, telecommunication appliances and transport vehicles that increase the quality of life (Ighodaro, 2010). No wonder Kemp (2005) describes electricity as the life breath of modern society. Without a safe, sustained, reliable and reasonably affordable supply of electricity to meet demand, a country can hardly make progress in its economic and social development. Examples of electricity’s role as a key input into economic activity are myriad. The primary sector of the economy, which includes activities related to extraction of resources from the earth (for example, mining and farming), uses electricity to operate heavy machinery that can make processes faster and more efficient, the secondary sector, which includes activities related to manufacturing and production, uses electricity to power factories, warehouses, and the equipment inside of them. The tertiary sector, which includes service business and i    retail operations, uses electricity for lighting, heating and cooling, and the operation of productivity tools like computers and printers (Fornelio, 2010). The experience of developed countries shows that the electricity supply sector played a crucial role in their economic development not only as a key input in their industrial, technological and scientific advancement but also as a key factor in improving the quality of life of their people. Even in poor countries, it is argued that the use of electricity is associated with improving the health and education standards of the poor. At the individual level, increased electricity use is likely to be one of the most important causes of improved welfare of the poor. At the national level, in the era of the digital economy, it is not easy to envisage development without the use of modern energy, particularly electricity (Wolde- Rufael, 2006). Though the electricity supply industry in Ghana contributes only 10 percent of its energy-supply mix, the industry is a key driver of economic growth and development, powering the country’s industrial, commercial and urban development. The industrial, agricultural, mining and services sectors of the Ghanaian economy, which together account for 75 percent of the country’s GDP, rely critically on the electricity industry for their survival. The electricity industry accounts for 14.7 percent of total energy share in the industrial sector, and 32.6 percent energy consumption in the formal manufacturing sector. It also contributes 2.8 percent of real GDP and 10.65 percent of industrial GDP. With a customer base of approximately 2 million, it is estimated that 45-47 percent of Ghanaians including 15-17 percent of the rural population have access to grid 2    electricity with a per capita electricity consumption of 358kWh while the average annual growth rate stands at 9.7 percent (Institute for Statistical, Social and Economic Research (ISSER), 2005). Electricity consumption also depends on the stage of economic growth. For the proponents of this argument, economic growth is a necessary condition to insure better standards of living. They believe that economic growth translates to higher wages and incomes for workers and more employment allowing them to purchase a big ticket item such as automobiles, houses, refrigerators, air- conditions, washing machines, mobiles, cooking utensils etc. which leads to higher electricity demand and hence electricity consumption (Shahbaz and Lean, 2011). According to Khan and Ahmed (2009), the heightened interest by the major economic powers at gaining a firm foothold on energy rich regions across the globe is an effort to insure their uninterrupted long term supply of energy for a seamless trajectory of economic growth path. Understandably, energy will still remain a major focus in the future and the battle for control will only intensify in the days ahead. Wolde-Rufael (2006) pointed out that the spurt in demand for electricity in emerging economics is closely linked with increases in income. The relationship between electricity consumption and economic growth has now become an important issue for academic research particularly regarding electricity policies. This is essential because when a nation is heavily dependent on electrical energy, environmental hazards could possibly have a negative impact on economic development. For instance, natural gas, normally used in producing 3    electricity, is a major contributor to CO2 emission. Coal alone accounts for over 50 percent of all CO2 emissions. Also, since the Earth Summit of Rio de Janeiro in 1992 and the Kyoto Protocol in 1997, which state that environmental degradation and climate changes are related to fossil energy consumption, some experts suggest a lowering of the world energy use (Eggoh, Rault & Bangaké, 2011). Therefore, whether the economic development takes precedence over electricity consumption or electricity itself is a stimulus for economic development has motivated curiosity and interest among economists and policy analysts over the past decade to investigate the relationship and direction of causality between electricity consumption and economic variables such as GNP, GDP, income, employment or electricity prices. This is because the direction of causality has significant policy implications (Jumbe, 2004). If a unidirectional causality running from economic growth to electricity consumption or if there is no causality in either direction, this signifies a less electricity dependent economy such that electricity conservation policies (for example, electricity rationing) would not affect economic growth. However, if the causality runs from electricity consumption to GDP, this signifies an electricity dependent economy such that electricity consumption is a stimulus for GDP growth, implying that shortage of electricity may negatively affect economic growth or may cause poor economic performance (Stern, 1999). Karanfil (2009) suggested that in exploring the causal links between electricity consumption and economic growth, economists should consider the 4    inclusion of other variables rather than just the bivariate case. According to him, the omission of relevant variable can seriously affect the result of the study. One of the variables, he suggested is the financial development variable such as the stock market, domestic credit to private sector and/or liquid liabilities. Statement of the Problem It cannot be denied that electricity is a necessity of daily life. It is pervasive in everything people do, from running their appliances to lighting their homes to fueling their entertainment and communications with the outside world. Electricity is also an indispensable factor for the social and economic development of societies and the usage levels of electricity is an indication of the economic prosperity of nations (ISSER, 2005). Over the past few decades, the role of electricity consumption in economic growth has attracted significant attention from energy and development experts and has been debated extensively. Even though it is very well known that there is a strong correlation between growth and electricity use, the issue of “causality” that is, whether economic growth leads to increases in electricity consumption or that electricity consumption is the engine of economic growth.- remains still to be answered (Masih & Masih, 1996). Empirical evidence especially time series studies have generally been mixed and inconclusive. Specific studies on the relationship between electricity consumption and economic growth in sub-Saharan African countries especially Ghana (Adom, 2011; Twerefo, Akoena, Egyir-Tettey & Mawutor, 2008; Akinlo, 2008; Lee, 5    2005; and Wolde-Rufael, 2006) are very few. Some of these studies have used cross-country regression method. This method is saddled with heterogeneity bias. As such, it fails to address the country-specific effects of electricity consumption on economic growth which may lead to inconsistent and misleading estimates (Akinlo, 2008; Lee, 2005 and Wolde-Rufael, 2006). In addition, some other studies for instance Adom (2011) and Twerefo et al. (2008) using time series employed a bivariate VAR framework, however, bivariate tests suffer from omitted variable problem and this may lead to erroneous causal inferences. It is against this background that this study seeks to investigate the relationship between electricity consumption and economic growth for Ghana using autoregressive distributed lag (ARDL) approach to cointegration. Objectives of the Study The general objective of this study was to investigate the relationship between electricity consumption and economic growth using annual time series dataset for Ghana. Specifically, the study sought to: 1. Investigate the long-run relationship between electricity consumption and economic growth. 2. Examine the short-run relationship between electricity consumption and economic growth. 3. Explore the nature of causal relationship between electricity consumption and economic growth. 6    4. To provide policy recommendations Hypotheses of the Study The study seeks to test the following hypotheses. 1. HO: There is no long run relationship between electricity consumption and economic growth. 2. HO: There is no short run relationship between electricity consumption and economic growth. 3. H0: There are no causal relationships between electricity consumption and economic growth in Ghana. Significance of the Study The research results are relevant to policy makers, development planners, finance experts and researchers. Investigating the electricity-growth nexus can be of principal interest to government and energy planning. It will help stakeholders formulate policies capable of enhancing the development and effectiveness of the energy system. Again, the results of the study can serve as a guide to constructing appropriate energy sector reforms and in evaluating the effectiveness of these reforms since countries undertaking reforms hope to achieve a more competitive, healthier, efficient and deeper energy system and also the results from these causality tests may provide vital information that would be of help in formulating 7    macro econometric models, effective and efficient energy policy to sustain the economic growth of the country. Also, there exist few studies using time-series analysis on the electricity consumption-economic growth relationship especially in the case of Ghana. This study adds to the existing literature. In so doing the study addresses some of the methodological issues inherent in the literature. Scope of the Study This study investigates the relationship between electricity consumption and economic growth in Ghana using annual time series data set for the period 1971 to 2010. The choice of the data coverage was informed by the fact that it was extremely challenging getting data below 1971 on electricity consumption per capita which is a key variable in the electricity consumption and economic growth relation. The study employed the following variables: Electricity consumption, economic growth, financial development, capital stock, labour force, inflation and dummy variable for economic reform and constitutional regime. Economic growth is proxied by changes in real gross domestic product (GDP) per capita and electricity consumption per capita is used to proxy electricity consumption. The variables such as capital stock, labour force, financial development and inflation were used to augment the model. Capital stock was proxied by the share of gross fixed capital formation to GDP and 8    financial development is proxied by M2 money supply as a percentage to GDP. Inflation is also measured by changes in the consumer price index. The study employs the recently developed Autoregressive Distributed Lag (ARDL) model otherwise known as the bounds testing approach to cointegration developed by Pesaran and Shin (1999) and Pesaran, Shin and Smith (2001). Organisation of the Study This study is organized into five chapters. Chapter one, which is the introductory chapter, presents a background to the study, problem statement, objectives of the study, hypotheses, significance and scope of the study as well as organisation of the study. Chapter two presents review of relevant literature, both theoretical and empirical on the relationship between electricity consumption and economic growth. Chapter three presents the methodological framework and techniques employed in conducting the study. Chapter four examines and discusses the results and main findings with reference to the literature. The final chapter presents the summary, conclusions and recommendations of the study. 9    CHAPTER TWO REVIEW OF RELATED LITERATURE Introduction The broad aim of this chapter is to present the review of related literature on the relationship between electricity consumption and economic growth in Ghana. This is aimed at getting supporting theories and empirical evidence for the study. The chapter is organised into three sections. The first section presents an overview of the Ghanaian economy with specific focus on the electricity industry. The second section presents and discusses the theoretical literature on the relationship between energy (electricity) consumption and economic growth and finally, the last section reviews the empirical literature on energy (electricity) consumption and economic growth. Overview of the Ghanaian Electricity Sector and Economic growth Electricity is produced and delivered to customers through three different steps; generation, transmission and distribution as demonstrated in Figure 1. First of all, before electricity can be delivered, it needs to be generated. Generation plants consist of one or more generating units, that is rotating turbines, which convert mechanical electricity into electricity. Through generation substations, which connect generation plants to transmission lines, the electricity is transferred 10    to the transmission system in order to be transported over long distances. The transmission system will make it possible to optimize the production of electricity within a country and for power trading between countries. A transmission substation contains transformers to step down voltage to distribution levels and is a kind of connection point between the transmission system and distribution system. The distribution system transfers and divides the electric power to be delivered to customers as electricity (Fornelio, 2010). Figure 1: The Electric Power Flow – from Generation to End Customers Source: Fornelio, 2010 In Ghana, these three-step processes are controlled by three different utility companies. The Volta River Authority (VRA) is a state-owned enterprise that is solely responsible for bulk power generation in the country. Currently VRA operates the Akosombo and Kpong hydro stations which happen to be the major power generation sources in the country. Limited generation is also 11    undertaken by a private company, the Takoradi International Company (TICo), a joint venture ship between VRA and CMS Electricity Inc. of the USA (ISSER, 2003). Ghana Grid Company (GRIDCo) is also responsible for transmitting power from bulk power plants to distribution lines while the Electricity Company of Ghana (ECG) and the Northern Electricity Department (NED), the latter being a directorate of the VRA are responsible for electric power distribution in the country to the final consumer. The Electricity Company of Ghana delivers power to customers in the southern half of the country comprising Ashanti, Western, Central, Eastern, Volta and Greater Accra Regions while the Northern Electricity Department has responsibility for supplying power to customers in the northern half of the country consisting of the Brong Ahafo, Northern, Upper East and Upper West Regions (Ghana Grid Company Limited, 2010). The Public Utilities Regulatory Commission (PURC) and the Energy Commission are two government agencies that regulate the utilities for the public good rather than private interests. The PURC is an independent body with primary responsibility for setting the tariffs that utilities charge their customers. The ECG on the other hand is tasked with licensing and regulating the technical operations of the utilities. Both regulatory agencies also ensure fair competition in the power market, enforce standards of performance for the provision of services to customers and protect both customer and utility interests. Electricity policy formulation is the preserve of the Ministry of Electricity while the Electricity 12    Foundation, a non governmental agency has been very active in promoting electricity efficiency measures (ISSER, 2005). The electricity sector has experienced significant growth over a decade now. In 1992, electricity and water sector recorded a growth rate of 12.02% which was 5.43% higher than the previous year. The primary reason, as reported in the budget statement and economic policy for 1993, included expansions in the national electricity grid under the rural electrification programme and the expansion and up-grading of some urban electricity distribution networks. In 2000, the sector witnessed a growth rate of 4.5% which was below the 1992 figure. In terms of the sectors relative contribution to total industrial growth in the country, the electricity sector contributed 10.21% of total industrial GDP in 2000. In 2005, the sector witnessed an increase in growth rate of 12.4% which translated into the sectors increased relative contribution to total industrial GDP of 11.9%. However, in 2007, the sector recorded a decrease in growth rate of 17.4% which caused the sector’s relative contribution to total industrial GDP to fall to 10.2%. (ISSER, 2007). The major reason behind the sectors decreased contribution was mainly due to the serious drought that thumped the Ghanaian economy in 2007 which led to plummet in the water level of Akosombo, the foremost power house for the country. 13    History of Electrical Power Generation in Ghana The history of electricity production in modern Ghana dates back during the colonial times in 1914 when the electricity supply sponsored by the government was initiated in Sekondi in the Western part of modern Ghana (ISSER, 2005). Since then various reforms and restructuring has taken place. However, it is not the focus of this section of the report to relate the long history and the various transition reforms that took place (thus, pre independence era). However the historical path of Ghana’s electrical power production can be divided into three main phases. a) Before the hydro years: this period refers to the time period before the main hydro plant in Akosombo was built in the 1966. b) The hydro years: refers to the time period from 1966 when the Akosombo hydro plant was completed to the 1980s. c) Thermal complementation years: the 1990s to present when thermal plants were used to supplement the hydro generation (Asare, 2008). This thesis focuses on the post independence era especially since the late 70s to the present time. Hydro Generation Currently, Ghana operates two main hydro power plants and two thermal plants. The first and the biggest hydro plant built is the Akosombo hydro plant with an installed capacity of 1020 MW located in the Eastern part of the country in 1966 and the main purpose initially was to supply electric power to the aluminum industry. The building of the Akosombo hydro dam flooded the Volta river basin creating the largest manmade lake in the world which covers 14    approximately 3.6% of Ghana’s land area. The power generated from the Akosombo plant serves as the driving force behind Ghana’s economic development and also supported neighbouring countries such as Togo and Benin by exporting power to these countries (Suave, Dzokoto & Opare, 2002). With Ghana’s expanding industry and the nature of economic development which caused greater demand for electric power in the 1980s, a second but smaller hydro power plant called the Kpong Hydro plant with an installed capacity of 160 MW was developed on the downstream of the same Volta River to supplement the Akosombo hydro plant. Between 1982 and 1984 occurred one of the most severe droughts in the Volta River Basin in recorded history (ISSER, 2005). This greatly affected the production output of the two hydro electric power plants and led to the search and additional sources of producing electricity other than hydro source. Thermal Addition To complement the existing hydro power plants, the Volta River Authority (VRA) in 1997 established the Takoradi Thermal Power Station (TPPS) in the Western region of the country, the first of its kind in Ghana. A 550 MW installed capacity with a joint private partnership as part of the government plans to allow private participation in the electricity generation sector. The Takoradi thermal Power Company consist of two companies all located in the same region. It consists of a 330 MW combined cycle plant called Takoradi Thermal Power Company (TAPCO) with a private partnership with CMS Electricity of USA in a 15    ratio of 10 percent (VRA) to 90 percent. And the second part is the Takoradi international company (TICO), A 220 MW installed gas turbine plants in a ratio of 10 percent (VRA) to 90 percent (Asare, 2008). Currently, all the thermal power plants are fuelled with light fuel. However with expected natural gas from Nigeria (as Ghana is part of a West African joint project called the West African Gas Pipe line project) high cost of fuel associated with the light fuel is expected to fall as natural gas is relatively cheaper than light fuel. There is also a 30 MW diesel power plant at Tema near the capital in the Greater Accra region of the country. Again a third hydro power plant called the Bui Dam with installed capacity of approximately 400 MW has been developed to complement the existing two hydro plants. Unlike the two other hydro plants, the Bui Dam is a joint venture between the Government of Ghana (GoG) and a Chinese construction company Sino Hydro (Asare, 2008). Electricity Demand and Supply in Ghana At present, Ghana’s electricity sector has a customer base of more than 2 million residential and commercial customers and 1,150 industrial customers. In 2009, these customers contributed to a peak power demand of 1,423 MW and a cumulative energy demand of 10,116 GWh (Ghana Grid Company, 2010). Peak demand is the maximum amount of electricity that customers consume instantaneously, while energy demand is the amount of electricity they use over time (thus, the sum of instantaneous demand over time). 16    Electricity demand in Ghana is divided across 40 load centres, which include cities, clusters of smaller towns and villages, and large industrial sites such as mines. In Ghana, a relatively small number of load centres account for a large fraction of total demand. In fact, Ghana’s ten largest load centres together accounted for nearly 68% of peak demand and 72% of energy consumption in 2009. Most of these load centres coincide with urban centres – Accra, Tema and Kumasi alone account for approximately 49% of total national peak demand (ISSER, 2007). The remaining major load centres are associated with heavy industrial activity. Industrial customers are characterized by very high, consistent power demands. The four largest mines alone accounted for 12.5% of national peak demand, and each mine consumed a significant amount of energy relative to their peak demand. For instance, the load centres at Takoradi and New Obuasi had similar peak demands of 44.7 MW and 53.0 MW while New Obuasi as a mining site had over 50% more annual energy demand than Takoradi (Ghana Grid Company, 2010). Generally, domestic consumption of electricity has been increasing continuously even though domestic production of electricity in Ghana has been inconsistent largely due to inadequate investment in additional capital and unpredictable nature of the weather. It is important to reiterate here that the balance between electricity demand and supply is an important prerequisite to ensuring a reliable electricity system. Ghana’s electricity sector has experienced considerable prolong power shortages over the last decade. For instance in 2000, 17    while electricity demand stood at 7488.9 GWh, electricity supply stood at 7233GWh representing a demand-supply gab of 265.9 GWh. Also in 2001, while domestic electricity stood at 8012.1 GWh representing a 7.1 percent increase, domestic electricity supply stood at 7857 representing an increase of 8.8 percent. This shows a demand – supply gab of 153.1 GWh. However, in 2008, while domestic electricity demand stood at 8066 GWh, domestic production of electricity stood at 8324 GWh representing a demand – supply gab of -257.8 GWh (Adom, 2011). Important to the analysis of the demand – supply gap is the existence of losses in the form of technical and non- technical losses of power within the electricity sector. Technical losses are largely caused by energy dissipated as heat in the restrictive, conductors and equipment used for transmission, transformation and distribution of power. Non-technical losses on the other hand include pilferage, defective metres and errors in accounting for electricity consumption. In practice, non-technical losses are largely confined to distribution while technical losses are present in generation, transmission and distribution (Adom, 2011). Generally, losses account for approximately 24% of electricity demand in Ghana which is normally driven by distribution losses (both technical and non- technical). In comparison, losses account for only 6.5% of demand in the United States of America (USA) and 21.21% in Rwanda. Transmission losses in Ghana are about 3.8 compared to an industry rule-of-thumb of 3% (Ghana Grid Company Limited, 2010). These huge losses in Ghana’s electricity sector are partially responsible for the current trend of deficit within the electricity sector. 18    One major step undertaken to curtail losses in the system has been the installation of prepaid metres. It must be said here that, although, this is a good step not much has been achieved in terms of national coverage. Environment and Energy Policy Issues Environmental concerns are a prominent part of every industry today and the electric power industry is no exception (ISSER, 2005). Coal and lignite are taken from underground and strip mines. Natural gas wells are drilled to provide fuel to generate electricity. Power plants that use fossil fuels emit pollutants that are subject to emissions regulations. Transmission lines also spread across the state, affecting human and natural environments. These activities are monitored and regulated, but because of the size and scope of the industry, there will always be concern about electric power and its environmental effects (Stern, 2003). According to Weedy (1987), power plants use various fuels that are linked to problems like acid rain, urban ozone depletion, global warming and waste disposal. Each fuel has its own environmental advantages and disadvantages. Coal, one of the lowest priced fuels, requires considerable treatment of emissions to meet environmental standards and its use triggers concerns about global warming. Natural gas, a more expensive fuel, burns cleaner than coal but can contribute to ozone formation in urban areas. Wind and solar power which require relatively high capital costs produce no direct emissions and have virtually no fuel cost but they can be unsightly or impact wildlife negatively. Nuclear power plants emit no combustion gases but have raised the issue of long-term disposal of spent fuel. 19    Ghana generates most of its power from hydroelectric facilities, which do not cause emissions of harmful elements into the atmosphere; but their large reservoirs have some impact on the environment by flooding large areas, causing people to move, changing their ecology and causing silt formation (Adom, 2011). He again argued that acid rain, urban ozone depletion, particulate emissions and global warming are the four primary air pollution concerns for the electric power industry. Power plants contribute relatively little to emissions of Volatile Organic Compounds (VOCs), carbon monoxide, nitrous oxide (a greenhouse gas and oxide of nitrogen), or methane (a greenhouse gas) in Ghana. Environmental concerns are also linked to energy issues such as the use of alternative fuels to generate electricity and energy efficiency. Alternative fuels are defined in different ways, but are often alternatives to conventional fuels. There are usually environmental or energy-related benefits associated with alternative fuels. For example, natural gas used as a transportation fuel is considered an alternative that can be cleaner than conventional gasoline, but natural gas for power generation is not considered to be an alternative fuel (it is considered by many to be a cleaner fuel). Alternatives for power generation include hydroelectric power, solar electricity, wind energy, biomass energy and geothermal power. The fuels for each of these are available at little or no cost but they are often renewable fuels and they may produce little in any direct emissions. Proponents argue that the environmental costs of conventional electric power are not reflected in the cost of electricity produced. If these costs (externalities) were included, alternative fuel electricity would be very competitive (ISSER, 2005). 20    Review of Theoretical Literature Reviewing the relationship between electricity consumption and economic growth is necessary to explain the causality between these variables. For this reason, theoretical literature relating to energy economics is examined. In this section, the theoretical literature in energy economics is discussed according to, the mainstream theory of growth and the causal relationship between energy consumption and economic growth. The Mainstream Theory of Growth Theoretical work on economic growth dates back many centuries ago. Although classical economists like Smith, Malthus and Ricardo did not formally study economic growth, however, they outlined the basic ingredients of economic growth. Recent neoclassical economists regarded as pioneers of economic growth analysed economic growth with rigorous models. They include Ramsey (1928), Solow (1956), and Swan (1956), while Romer (1986) and Lucas (1988) initiated economic growth theories that are known as endogenous growth theories. However, there is an inbuilt bias in mainstream growth theory that down play the role of resources in the economy, though there is nothing inherent in economics that restricts the potential role of resources in the economy. The Basic Growth Model The most simple growth model which examines the hypothetical economy is the model developed by Solow (1956) and Swan (1956), which is referred to as 21    the neoclassical growth model. The model is assumed to be a function of capital, labour and technology. The critical assumption of the model is that the production function has constant returns to scale, exhibits positive and diminishing marginal products with respect to each input and satisfies the properties that the marginal products of inputs approaches infinity as inputs go to zero and approaches zero as inputs go to infinity. By assuming a constant rate of saving and diminishing returns to capital, the model predicts that growth in the long-run is a function of only technical change and not of savings or investment. According to the Solow model, saving will have an effect on the level of income but not on its growth rate. This prediction implies that in the absence of continuous improvement in technology, per capita growth will eventually cease (Stern, 2003). Another important neoclassical growth theory, beside the Solo-Swan model, is the Ramsey model. The Ramsey model is a refinement of the Solow- Swan model. Chronologically, Ramsey’s growth theory was developed before the Solow-Swan’s. In the literature, however, his model is usually put after the Solow-Swan’s. One of the key features in his model is the assumption that households optimise their utility over time. This assumption essentially makes the model dynamic. Using Ramsey’s model as their starting point, Cass (1965) and Koopmans (1965) recast the saving rate that is exogenous under Solow-Swan model as endogenous. Even though this is considered a refinement of the neoclassical growth model, it does not eliminate the dependence of the long-run growth rate on exogenous technological progress (Solow, 1956). The works of 22    Cass and Koopmans (1965) actually mark the end of the basic neoclassical growth era. The simple growth model described does not explain how improvements in technology come about. The model just assumed that it happens exogenously, so this model is said to have exogenous technological change. More recent models attempt to endogenize technological change by explaining technological progress as the outcome of decisions taken by firms and individuals. It was first started by the work of Romer (1986) and Lucas (1988). Unlike the Solow-Swan model in which the long-run growth rate is determined by an exogenous technological progress, Romer and Lucas suggest a model that endogenises the growth rate of the economies. Models that have endogenous source of per capita economic growth are called the endogenous-growth models. Endogenous growth models are similar to the neoclassical models but they differ considerably in their underlying assumptions and suggested conclusions. Three distinctions between the two are obvious: First, the neoclassical assumption of diminishing marginal returns to capital is discarded. Second, the model also envisages increasing returns to scale in aggregate production, and third, the model recognizes the role of externalities in determining the rate of return on capital. In endogenous growth models, the relationship between capital and output can be written in the form Y = AK. Where A can be interpreted as representing any factor that affects technology, while K represents both human and physical capital. Unlike the neoclassical model, the endogenous growth theorists consider 23    the term A in the expression as a constant, and so growth can continue indefinitely as capital is accumulated (Romer, 1956). According to the endogenous growth model, technological knowledge is thought of as a form of capital and it is accumulated through research and development (R&D) and other knowledge creating processes. A firm increase technological knowledge through investment in capital. Each firm‘s technological knowledge has two special properties. First it is a public good such that the stock of it is not depleted with use and more is available for others at zero cost and secondly, it generates positive externalities in production. This means that whilst the firm doing R&D obtains benefits from the knowledge acquired, others benefit too –indicating that the benefits that the firm accrues when it learns and innovates are only partly appropriated by itself. There are beneficial spillovers to the economy from the R&D process so that the social benefits of innovation exceed the private benefits to the original innovator (Lucas, 1988). So in an endogenous growth model, technological knowledge through investment in capital exactly offset the diminishing returns to manufactured capital and the economy can sustain a constant growth rate. So growth in this case is permanently influenced by the savings rate therefore a higher savings rate increases the economy’s long run growth rate. Growth Models with Resources and no Technical Change The mainstream theory of economic growth discussed pays little attention to the role of energy in economic growth. However, adding non renewable natural 24    resources that are essential in production to the basic mainstream growth models means that capital also needs to be accumulated to compensate for resource depletion. All natural resources exist in finite quantities. Some of the resources are non-renewable and many renewable such as energy are also finite. Finiteness and exhaustibility of resources poses problems especially to economic growth (Stern, 1999). The neoclassical literature about economic growth is mainly concentrated on which conditions permit continuing growth. Technical and institutional however determine whether growth is continous or not. Technical conditions include the mix of renewable and non renewable resources, the initial endowments of capital and natural resources, and the ease of substitution among inputs. The institutional setting on the other hand includes market structure (competition versus central planning), the system of property rights (private versus common property, etc.), and the system of values regarding the welfare of future generations (Stern, 2003). When the natural resources are exhausted, they are replaced by their substitutes or by equivalent forms of human-made capital (people, machines, factories, etc.) for production. Neoclassical economists concern themselves with what institutional arrangement provide for continuing growth, on the other hang they ignore the impact of technical arrangement. That is, they only assume that substitutability is achievable in a model with only a nonrenewable natural resource with no extraction costs and non depreciating capital when the elasticity 25    of substitution between the two inputs is unity, and when certain other technical conditions are met (Solow, 1956). To the neoclassical, though substitutability is technically feasible, they were only interested in which institutional arrangements provide continuing economic growth. Growth Models with Resources and Technical Change In addition to substitution of capital for resources, technological change might permit growth or at least permit constant consumption in the face of a finite resource base. Stiglitz showed that, when the elasticity of substitution between capital and resources is unity, exogenous technical progress will allow consumption to grow over time if the rate of technological change divided by the discount rate is greater than the output elasticity of resources (Stiglitz, 1974). The increase in total productivity technically increases the ease of substitutability of economy and substitutability can enable technological change even with an elasticity of substitution of less than one. However, technical feasibility does not mean that there will be sustainability. Technical improvement means increase in the quantity of output for each unit of inputs. Studies that examine the role of resource in growth with endogenous technical change such as Smulders and de Nooij (2003), and Aghion and Howitt (1998) have been less than the studies with the exogenous technology change or no technological change assumptions. Studies with endogenous technological 26    change have not yet provided conditions for the achievement of substitutability (Stern, 2003). In a study by Aghion and Howitt (1998) about the role of natural resources in production, they authors used four different models to examine whether sustainable growth will be achieved or not. First two models cover renewable resources, the other two cover non-renewable resources. In the models which cover non-renewable resources, it is assumed that non-renewable resources are important for production. Conversely, in the models which cover renewable resources, the resource which decreases the amount of pollution of the environment has more importance than non-renewable ones. Tahvonen and Salo (2001) also developed a model which covers renewable and non-renewable energy resources at the same time. This model is more realistic than the previous neoclassical approach. They intend to see how the growth process would actually work. The extraction costs for fossil fuels and the costs of production for renewable resources are included in the model. This model also investigated the situation in which there is no technological change, thus, technological change is assumed to be exogenous. The study also assumed that increase in extraction leads to increase in technological knowledge and that technological knowledge increases the capital stock. The optimal development in such an economy appears to mimic history much more effectively than the neoclassical models. When the economy is divided by pre-industrial, industrial and post-industrial eras, fossil fuels usage rises in the first two eras and then falls 27    as capital accumulation rises. The price of non-renewable energy resources first falls and then rises. Ayres and van den Bergh (2005) also proposed a more disaggregated view on growth engines or mechanisms. They offered a model of economic growth with energy resources and dematerialization, and considered three growth mechanisms: firstly, the resource use (fossil fuel) growth engine, secondly, the scale-cum-learning growth mechanism and finally, the value creation (dematerialization) growth engine. They concluded that for sufficiently high growth rates, the required resource input must increases linearly with income. They argued that although theoretical results provide insufficient information on future patterns of growth in relation to resource use, the relevant policy tool is R&D investment, supplemented by regulation as applied to natural resource utilization, especially energy use efficiency and dematerialization. Smulders and de Nooij (2003) further argued that energy use has a positive growth rate apart from a possible on-time reduction in the level of energy use. To them, the level of technology affects the use of energy, while the availability of investment capital has considerable impact on the energy consumption and economic growth. Energy Consumption and Economic Growth The study of the empirical investigations into the causal relationships between energy consumption and economic growth can be analysed through two lines; the hypothesis criteria (Apergis and Payne, 2009) and the generation criteria 28    (Guttormsen, 2004). The hypothesis approach analyses the causation in light of whether studies concluded that energy (electricity) consumption causes economic growth or otherwise or both. Along these lines, studies on the empirical investigation into the energy (electricity) and economic growth nexus have been grouped into four; the Growth-led-Energy hypothesis, the Energy led-Growth hypothesis, the Energy-led-Growth-led-Energy hypothesis, and the neutrality hypothesis (Adom, 2011). The Growth-led-Energy hypothesis asserts that economic growth leads to energy consumption. This implies that even severe energy crisis will not retard economic growth, hence energy conservation measures are a viable option. The Energy-led-Growth hypothesis asserts that energy consumption leads to economic growth. This suggests that severe energy crisis will retard economic growth, hence energy conservation measures are not a viable option. The Energy-led- Growth-led-Energy hypothesis asserts that there exists bidirectional causality between energy consumption and economic growth. Lastly, the neutrality hypothesis asserts that there is no causal relationship between energy consumption and economic growth (Apergis and Payne, 2009). Along the lines proposed by Guttormsen (2004), studies on the empirical investigations into energy and economic growth have been classified along three lines; the first generation studies, the second generation studies, and the third generation studies. The first generation studies consist of studies that basically used the traditional Vector Autoregressive Models and the standard Granger causality test. The main weakness associated with this generation of studies is that 29    they assume the series to be stationary. As a result the second generation of studies proposed cointegration as the appropriate tool to use in analyzing the causal relationship between energy consumption and economic growth. Thus, in the second generation of studies, pairs of variables were tested for cointegration relationship and an error correction model was estimated to test for causality (Engle and Granger, 1987). However, given the possibility of more than one cointegrating vectors, the second generation studies approach was deemed inappropriate (Granger, 1988). This led to the third generation of studies, which proposed a multivariate approach that allowed for more than two variables in the cointegrating relationship. This approach facilitates estimations of systems where restriction on cointegrating relationship can be tested and information on short-run adjustment can be investigated. There are two main problems with the third generation studies. First, the third generation studies impose restrictions that the variables should be integrated of order one. Secondly, the variables will have to be cointegrated before a test of causality can be possible. This has led to the fourth generation of studies. These studies use the Toda and Yomamoto Granger Causality test, which is based on the Autoregressive distributed lag model. In this generation of studies, restrictions are not imposed on the variables. Thus, causality is still possible even when variables are integrated of order zero, one or both. In other words, this approach allows for the test of causality even when variables are not cointegrated (Adom, 2011). 30    In addition to the above, Ozturk and Acaravci (2010) in a literature survey on the energy-growth nexus classified the various studies into country-specific and multi-country studies on energy (electricity consumption) and economic growth. The general observation according to this study is that the results emanating from the multi-country studies and country-specific studies on the causality between energy consumption and economic growth reveals contradictory results. However, the results from the country-specific studies on the causality between electricity consumption and economic growth reveals that there exists a positive causality which runs from electricity consumption to economic growth but the multi-country studies on the causality between electricity consumption and economic growth shows contradictory results. Review of Empirical Literature The causal relationship between electricity consumption and economic growth has been a debated subject of the extensive empirical literature in the past three decades. However, no common consensus neither on the existence nor on the direction of the causal relationship between electricity consumption and economic growth has emerged. This depends on institutional, structural and policy differences of the countries involved, variety of variables and data span chosen and methodological differences. The aim of this section is to summarize the empirical literature of causality relationship between electricity consumption and economic growth and present the inconsistency of these studies. 31    The pioneering study by Kraft and Kraft (1978) using Sims’s technique found evidence of unidirectional causality, only running from gross national product (GNP) to energy (electricity) consumption for the United States (US) over the period 1947-1974. Therefore the US economy as at that time frame is not energy (electricity) dependent. It implies that policies aimed at reducing electricity consumption can be implemented with little adverse or no effect on economic growth. On the other hand, Akarca and Long (1980) showed that there is no causal relationship between energy consumption and economic growth when the sample time period is shortened by 2 years in the US economy. The techniques used to analyze the dynamic relationships between these two economic series in the study included cross-correlation functions based on the double-filter and single-filter, the haugh test of independence, the Sim's test of causality and the generalized Box-Jenkins procedure of multiple time series modeling. The study concludes that energy conservation does not lead to an increase or decrease in total employment. Yu and Hwang (1984) also found evidence in support of the neutrality hypothesis that there is no causal relationship between energy consumption and GNP in US using the Sims’ and Granger causality test when the sample period is extended by 5 years. It implies that energy consumption is not correlated with GNP, so that energy conservation policies do not affect GNP. In a similar study, Yu and Choi (1985) examined the causal linkage between GNP and the aggregate and as well as several disaggregate categories of energy consumption including solid fuels, liquid fuels, natural gas, and electricity for five countries with various 32    stages of economic development for the time period 1950-1976 based on Sims and Granger tests of causality. This study indicates unidirectional causality from aggregate energy consumption to GNP for the Philippines and from GNP to aggregate energy consumption for South Korea but no causality in either direction for the USA, the United Kingdom (UK), and Poland is found. If causality only runs from energy consumption to GNP, then it implies that the economy is energy dependent and the shortage of the energy may negatively affect economic growth. Erol and Yu (1987) also use the results of the Sims and Granger causality tests between energy consumption and GNP in some industrialized countries for the period 1950-1982 to conclude a unidirectional causality running from energy consumption to GNP for Canada, from GNP to energy consumption for both West Germany and Italy, neutrality of energy consumption with respect to GNP for France and UK and a bidirectional causality in Japan. Stern (1993) uses a multivariate approach rather than a bivariate approach to examine the Granger causality between GDP and energy consumption using vector autoregressive (VAR) model of GDP, energy use, capital stock, and employment for the period 1947-1990 in USA. Stern uses GDP instead of GNP and a quality-adjusted index of energy input rather than gross energy use different from many previous studies. As Glasure and Lee (1997) argue that the use of GDP is better than the GNP since the country’s total energy consumption depends on goods and services produced within the country, not outside the country. Stern (1993) finds that energy consumption Granger causes GDP. This result 33    contradicts the previous results for USA and the changes may probably be caused by variation in the variables and in the time span. Ebohon (1996) tests the Granger causality between electricity consumption and economic growth that is proxied by GDP and GNP for two countries, Tanzania over the period 1960 to 1984 and Nigeria over the period 1960 to 1981. The result of the study reveals bidirectional causality between electricity consumption and economic growth for both nations. This means that, unless energy supply constraints are eased, economic growth and development will remain elusive to these countries. The finding of the study supports the view that energy consumption plays an important role in the economic development of countries. Nachane, Nadkarni and Karnik (1998) first use the cointegration theory to test the existence of the long run equilibrium relationship between energy consumption and economic growth using GDP as a proxy. The study also shows the strength of the causal relationship. Conversely, previous studies only indicate the causal relationship between the variables. Nachane et al. (1998) study the relationship between energy consumption and economic growth for 25 countries for the period 1950 to 1985 but cointegration can only be established for 16 countries (11 developing and 5 developed countries). Bidirectional causality is found by using the Sims’ and Granger causality tests for all countries except Colombia and Venezuela. Yu and Jin (1992) also examined the bivariate cointegration between energy consumption and income or employment using monthly USA data over 34    the period 1974 to 1990. They found that cointegration fails to exist between them for either energy consumption-income or energy consumption-employment relationship by using Engle-Granger two step procedure. This implies that energy consumption is neutral with regard to income and employment over the long-run. This analysis is consistent with the earlier conclusions in the literature for the economy that energy consumption policies do not affect the growth in the short run. Stern (2000) extends his previous study of USA in the post-war period by adding multivariate cointegration relationship between energy consumption and GDP. The results show that cointegration does occur and that energy input cannot be excluded from the cointegration space. Stern concludes that there is a unidirectional causality from energy consumption to GDP. This means that energy consumption is a limiting factor for economic growth. This conclusion is similar to multivariate model of Stern (1993), while it contradicts the bivariate model of Yu and Jin (1992). In a similar study, Ghosh (2002) investigated the Granger causality between electricity consumption per capita and Gross Domestic Product (GDP) per capita for India using annual data covering the period 1950–51 to 1996–97. The Phillips–Perron tests revealed that both series, after logarithmic transformation, are non-stationary and individually integrated of order one. The study found no strong evidence of long-run equilibrium relationship among the variables but there exists unidirectional Granger causality running from economic 35    growth to electricity without any feedback effect. Ghosh therefore suggested that electricity conservation policies can be initiated. Guttormsen (2004) also applied the causality test to empirically examine the causal relationship between primary energy consumption and real Gross National Product for Turkey during 1970–2006. The study employed unit roots tests; the augmented Dickey–Fuller (ADF) and the Philips–Perron (PP), Johansen cointegration test, and the Pairwise Granger causality test to examine the relationship between the series. The empirical results from the study indicated that the two series are non-stationary, however the first differences of the series led to stationarity. Further, the results indicated that energy consumption and GNP are cointegrated and there is bidirectional causality running from energy consumption to real GNP and vice versa. This means that an increase in energy consumption directly affects economic growth and that economic growth also stimulates further energy consumption. Also, Akinlo (2009) examined the association between energy consumption proxied by electricity consumption and real GDP for the case of Nigeria during the period 1980-2006. The empirical evidence shows cointegration for both variables and electricity consumption Granger causes real GDP. The results of Hodrick–Prescott (HP) filter used by the study also decompose the fluctuations from the series of electricity consumption and economic growth. The estimation results showed that there is cointegration between the trend and the cyclical components of the two series, which seems to suggest that the Granger causality is possibly related with the business cycle. The paper therefore 36    suggested that investing more and reducing inefficiency in the supply and the use of electricity can further stimulate economic growth in Nigeria. Odhiambo (2009) examined the causal relationship between electricity consumption and economic growth in South Africa covering the years 1971-2006. The study incorporated the employment rate as an intermittent variable in the bivariate model between electricity consumption and economic growth, thereby creating a simple trivariate causality framework. The empirical results showed that there is bidirectional causality between electricity consumption and economic growth in South Africa. In addition, the results showed that employment in South Africa Granger-causes economic growth. It is observed from the study that the results apply irrespective of whether the causality is estimated in the short-run or in the long-run formulation. The study, therefore, recommends that policies geared towards the expansion of the electricity infrastructure should be intensified in South Africa in order to cope with the increasing demand exerted by the country's strong economic growth and rapid industrialization programme. Odhiambo’s findings of a bidirectional relationship between electricity consumption and economic growth parallel the results obtained by Jumbe (2004) for Malawi. Belloumi, (2009) in a related study used the Johansen cointegration technique to examine the causal relationship between per capita energy consumption and per capita gross domestic product for Tunisia during the 1971– 2004 period. In order for the study to test for Granger causality in the presence of cointegration among the variables, a vector error correction model (VECM) is 37    used instead of a vector autoregressive (VAR) model. The estimation results indicated that the per capita gross domestic product and per capita energy consumption for Tunisia are related by one cointegrating vector and that there is a long-run bidirectional causal relationship between the two series and a short-run unidirectional causality from energy to gross domestic product. Similarly, Ighodaro (2010) reexamined cointegration and causality relationship between energy consumption and economic growth for Nigeria using data covering the period 1970 to 2005. Unlike previous related study for Nigeria, different proxies of energy consumption (electricity demand, domestic crude oil consumption and gas utilization) were used for the estimation. The study included government activities proxied by health expenditure and monetary policy proxied by broad money supply though emphasis was on energy consumption. Using the Johansen co-integration technique, the study found that there exist a long run relationship among the series and all the variables used for the study were found to be I(1). Furthermore, unidirectional causality was established between electricity consumption and economic growth, domestic crude oil production and economic growth as well as between gas utilization and economic growth in Nigeria. While causality runs from electricity consumption to economic growth as well as from gas utilization to economic growth, it was found that causality runs from economic growth to domestic crude oil production. The result of unidirectional causality from electricity consumption to economic growth is consistent with the result obtained by Akinlo (2009). 38    Masih and Masih (1996) used the cointegration results between energy consumption and economic growth in testing Granger causality for Asian countries. Long-run energy and income relationship is only held for India, Pakistan, and not for Malaysia, Singapore and the Philippines. With the aid of cointegration and VECM, they found unidirectional causality from energy consumption to GDP for India, exactly the reverse for Indonesia and bidirectional causality for Pakistan. On the contrary, they did not find any direction of causality between these two variables by applying VAR model for the three non- cointegrated countries. Glasure and lee (1997) examined the causal relationship between energy consumption and GDP for South Korea and Singapore over the period 1961 to 1990 by using cointegration and VECM and VAR-based standard Granger causality test. The result of VAR indicated no causality between energy consumption and GDP for South Korea and unidirectional causality from energy consumption to GDP for Singapore. In contrast, the result of VECM revealed bidirectional causality between energy consumption and GDP, because standard Granger causality test is not able to estimate long run relationship. Masih and Masih (1997) reexamined the causality analysis of energy consumption and economic growth based on the demand side multivariate model. The study used trivariate variables of energy consumption, GDP, and consumer price index as a proxy for real energy price rather than bivariate system of energy consumption and GDP. The trivariate model applied in the study is different than the production side model which consisted of energy consumption, GDP, capital, 39    and labour in Stern’s (1993) studies. Their cointegration and VECM results indicated that there exists a long-run equilibrium relationship among energy consumption, GDP, and price and bidirectional causality between energy consumption and GDP for both in South Korea and Taiwan. Furthermore, Masih and Masih (1998) used multivariate cointegration and error correction modeling techniques to estimate the causal relationship for two Asian less-developed countries: Thailand and Sri Lanka. They discovered that energy consumption, GDP, and price are cointegrated and there is a unidirectional causality from energy consumption to GDP and price. Therefore energy consumption is relatively exogenous variable in these countries. Cheng (1999) uses Granger causality, cointegration and error correction approach for India during 1952-1995. The cointegration test revealed that energy consumption, GNP, capital and labour are cointegrated and economic growth unidirectionally Granger causes energy consumption both in the short run and in the long run. The result of the study contradicts with Masih and Masih’s (1998) results by finding unidirectional causality from energy consumption and economic growth for Thailand and Sri Lanka. Variables and time period are different in these studies. Asafu-Adjaye (2000) in a similar study tested the causal relationship between energy consumption and GDP by using a model based on demand functions that includes energy consumption, GDP, and price in four Asian developing countries (India, Indonesia, the Philippines and Thailand). The results indicated that, in the short-run, unidirectional Granger causality runs from energy to income for India and Indonesia, while bidirectional Granger causality runs 40    from energy to income for Thailand and the Philippines. In the case of Thailand and the Philippines, energy, income and prices are found to be mutually causal. The study results did not support the view that energy and income are neutral with respect to each other, with the exception of Indonesia and India where neutrality was observed in the short-run. Asafu-Adjaye’s results for Indonesia, the Philippines, and India are different from the result of Masih and Masih (1996) and Cheng (1999). However Asafu-Adjaye’s result for India is consistent with the finding of earlier study done by Masih and Masih (1996). Glasure (2002) applies five variable VECM consisting of real money supply a proxy for monetary policy, real government expenditure a proxy for government activity, dummy variable for the two oil price shocks, real oil price, energy consumption, and GDP to investigate the link between energy consumption and GDP for Korea by using VECM. He uses yearly data for the period 1961-1990 to conclude a bidirectional causality running between energy consumption and GDP. This result for Korea is similar to results of the Masih and Masih (1997) and Glasure and lee (1997) studies which show the bidirectional causality between energy consumption and economic growth. In a related study, Shiu and Lam (2004) employed Johansen and Juselius cointegration and short run Granger causality approaches for cointegration to examine the direction of causality between electricity and GDP per capita for China for the period 1978–2004. The estimation results indicated that real GDP and electricity consumption for China are cointegrated and there was only unidirectional Granger causality running from electricity consumption to real 41    GDP but not the vice versa. Then Hodrick–Prescott (HP) filter was applied to decompose the trend and fluctuation component of the GDP and electricity consumption series. The estimation results indicated that there is cointegration between not only the trend components, but also the cyclical components of the two series, which implies that, the Granger causality is probably related with the business cycle. Moreover, Jumbe (2004) applies Granger causality test, cointegration and error correction approach for discussing the relationship between various kinds of GDP, including overall GDP, agricultural GDP, and non-agricultural GDP and electricity consumption for Malawi during 1970-1999. The Granger causality results indicated bidirectional causality between electricity consumption and overall GDP and a unidirectional causality relationship running from non-agricultural GDP to electricity consumption. On the other hand, the VECM results showed a unidirectional causality from overall GDP and non- agricultural GDP to electricity consumption Oh and Lee (2004a) employed cointegration and vector error correction modeling techniques to estimate the causal relationship between energy consumption and economic growth in Korea for the period 1970-1999. By applying multivariate module of energy consumption, GDP, capital, and labour, they found unidirectional causality running from energy consumption to economic growth in the short run and bidirectional causality between energy consumption and GDP in the long run. Oh and Lee (2004b) again used quarterly data over the period 1981-2000 for Korea. They applied cointegration and VECM on two multivariate models: one a demand side model and the other a production 42    side model. In their results, VECM shows no causal relationship for energy consumption and GDP in the short run but bidirectional causal relationship between energy consumption and GDP in the long run. When the time period is changed, they find contradictory causality results for Korea. Paul and Bhattacharya (2004) empirically examined the relationship between electricity consumption and economic growth using monthly India data from 1950 to 1996. Their empirical results revealed a short-run unidirectional causality from energy consumption to GDP using standard Granger causality test, a long-run unidirectional causality form GDP to energy consumption using Engle- Granger cointegration approach and a short-run unidirectional causality from GDP to energy consumption using Johansen cointegration approach. The results of standard Granger causality test combine with the Engle-Granger cointegration approach are same as the results of Johansen cointegration approach. Similarly, Jamil and Ahmad (2010) empirically examined the relationship among electricity consumption, its price and real GDP at the aggregate and sectoral level in Pakistan. Using annual data for the period 1960–2008, the study found the presence of unidirectional causality from real economic activity to electricity consumption. In particular, growth in output in commercial, manufacturing and agricultural sectors were found to increase electricity consumption, while in residential sector, growth in private expenditures was the cause of rising electricity consumption. The study concludes that electricity production and management needs to be better integrated with overall 43    economic planning exercises. This is essential to avoid electricity shortfalls and unplanned load shedding. Twerefo et al (2008) also utilized the technique of cointegration and vector error-correction modeling to examine the link between electricity consumption and economic growth in Ghana. The results suggested that GDP in Ghana indeed Granger causes electricity consumption and that the causality is unidirectional. The study advocated that electricity conservation is a viable option for Ghana. Akinlo (2008) also explored the causal relationship between energy consumption and economic growth for eleven Sub-Saharan African countries. Using the autoregressive distributed lag (ARDL) bounds test, the study found that energy consumption is cointegrated with economic growth in Cameroon, Cote D'Ivoire, Gambia, Ghana, Senegal, Sudan and Zimbabwe. Moreover, the test suggested that energy consumption has a significant positive long run impact on economic growth in Ghana, Kenya, Senegal and Sudan. Granger causality test based on the vector error correction model (VECM) showed that there is bidirectional relationship between energy consumption and economic growth for Gambia, Ghana and Senegal. However, Granger causality test revealed that economic growth Granger causes energy consumption in Sudan and Zimbabwe. The neutrality hypothesis is confirmed in respect of Cameroon and Cote D'Ivoire. The same result of no causality was found for Nigeria, Kenya and Togo. The result showed that each country should formulate appropriate energy conservation policies taking into cognizance of her peculiar condition. 44    Cheng (1995) studies the temporal causal relationship between energy consumption and economic growth for the USA in 1947-1990 based on both bivariate and multivariate model. Case of non-causality is found in the USA using Hsiao’s version of the Granger causality for both bivariate model of energy consumption and GDP and multivariate model of energy consumption, GDP, and capital. Test of the causal relationship between energy consumption and economic growth in three Latin American countries applying Hsiao’s version of the Granger causality is the subject of the study by Cheng (1997). His results indicated no causality in either direction in Mexico for the 1949-1993 periods and in Venezuela for the 1952-1993 period in a multivariate model. However, in Brazil, for the 1963-1993 period, Cheng (1997) revealed a unidirectional causal relationship running from energy consumption to GDP without feedback in his bivariate model. In addition, capital was found to negatively, though weakly, causes economic growth for both Mexico and Venezuela. Cheng and Lai (1997) empirically examined the causality between energy consumption and economic growth in a bivariate model for Taiwan over the period 1955 to 1993. The Phillips-Perron tests employed in the study revealed that the series with the exception of GNP are not stationary and therefore differencing is performed to secure stationarity. They found no cointegrating relationship between energy consumption and GDP. The results of Hsiao’s version of the Granger causality methodology indicated that there is unidirectional causality from GDP to energy consumption without any feedback in Taiwan. Yang (2000) 45    also tested the causal relationship between income and various kind of energy consumption in Taiwan using 1954-1997 data. Using cointegration and Hsiao’s version of the Granger causality, the estimation results indicated bidirectional causality between energy consumption (total, coal and electricity) and GDP, unidirectional causality running from GDP to oil consumption and unidirectional causality from gas consumption to GDP. His result for total energy consumption and GDP does not support the previous finding of Cheng and Lai (1997) of unidirectional causal relationship running from GDP to energy consumption. Cheng (1998) again applied the Hsiao’s version of Granger causality test to investigate the causality between energy consumption and GNP for Japan in 1952-1995 based on production side model of energy consumption, GNP, employment and capital. The study found employment, energy consumption, Real GNP and capital not cointegrated. However, electricity consumption was found to negatively Granger causes employment whereas employment and real GDP are found to directly Granger causes electricity consumption. It is also found in the study that capital negatively Granger causes employment while Real GNP and employment are found to strongly influence electricity consumption. The findings of this study seem to suggest that a policy of energy conservation may not be detrimental to a country such as Japan. In addition, the finding that energy and capital are substitutes implies that energy conservation will promote capital formation, given that output is constant. Aqeel and Butt (2001) also investigated the causal relationship between energy consumption and economic growth and energy consumption and 46    employment in Pakistan for the period 1955 to 1996. By applying techniques of cointegration and Hsiao’s version of Granger causality, the study concluded that there is no cointegration between the variables and unidirectional causality is found running from GDP to total energy and oil consumption and unidirectional causality running from electricity consumption to GDP. Altinay and Karagol (2004) empirically examined the causal relationship between electricity consumption and real GDP in Turkey during the period of 1950–2000 by applying bivariate model. Their study indicated that energy consumption and GDP are stationary series with different structural breaks by the Zivot and Andrews test and there is no evidence of causality relationship between energy consumption and GDP using Hsiao’s version of the Granger causality test. Yoo (2006) also employed cointegration and Hsiao’s version of the Granger causality tests to empirically examine the causality relationship between electricity consumption and economic growth in the four countries of the association of south East Asian nations (ASEAN), namely Indonesia, Malaysia, Singapore, and Thailand by using data for the period 1971 to 2002. Evidence from the study showed that there is no cointegration relationship between electricity consumption and GDP for all countries but there is a bi-directional causality between electricity consumption and economic growth in Malaysia and Singapore. This means that an increase in electricity consumption directly affects economic growth and that economic growth also stimulates further electricity consumption in the two countries. However, unidirectional causality was found 47    running from economic growth to electricity consumption in Indonesia and Thailand without any feedback effect. Chontanawat, Hunt and Pierse (2006) used the cointegration and Hsiao’s version of Granger causality methodology to analyse the causal relationship between energy consumption and output for 30 OECD countries over the period 1971-2000 and 78 non-OECD countries over the period 1971-2000. This is actually the first of such a large number of countries. Causality from aggregate energy consumption to GDP and GDP to energy consumption is found to be more prevalent in the developed OECD countries compared to the developing non- OECD countries; implying that a policy to reduce energy consumption aimed at reducing emissions is likely to have greater impact on the GDP of the developed rather than the developing world. Chebbi and Boujelbene (2008) in a similar study reinvestigated the cointegration and causality link between energy consumption and agricultural and non-agricultural outputs. In the study, ADF and KPSS, Johansen and VECM methods are used for 1971-2003 period in Tunisia. Empirical results suggest that there is only unidirectional causality running from agricultural and non- agricultural sectors to energy consumption. This unidirectional causality signifies a less dependent energy dependent economy for Tunisia. Fatai, Oxley and Scrimgeour (2001) empirically examined the causal relationship between GDP and various categories of energy consumption in New Zealand for the period 1960-1999 and the results are compared with Australia and several other Asian economies. They used the standard Granger causality and a 48    modified version of Granger causality proposed by Toda and Yamamoto for New Zealand and the results are similar with both methodology. They found unidirectional causality running from GDP to industrial and total energy consumption. The standard Granger causality test, the Toda-Yamamoto approach and the ARDL approach were also applied for Australia. They discovered unidirectional causality running from GDP to coal, electricity and final energy consumption using the standard Granger causality test and the Toda-Yamamoto test. On the other hand, the result for coal consumption and GDP is inconclusive for the ARDL approach. From the results, energy conservation policies may not have significant impacts on real GDP growth in industrialized countries such as New Zealand and Australia compared to the other Asian economies. Wolde-Rufael (2004) also investigated the causal relationship between various kinds of industrial energy consumption and GDP in Shanghai for the period 1952–1999 using a modified version of the Granger (1986) causality test proposed by Toda and Yamamoto. The empirical evidence from the disaggregated energy series suggested that there was a unidirectional Granger causality running from coal, coke, electricity and total energy consumption to real GDP but no Granger causality running in any direction was found between oil consumption and real GDP. Wolde-Rufael (2005) again studied the causal relationship between energy consumption and economic growth for 19 Africa countries between 1971 and 2001. The Toda-Yamamoto test results indicated a bidirectional causality for energy consumption and GDP for Gabon and Zambia, a unidirectional causality 49    from GDP to energy consumption for Algeria, Democratic Republic of Congo, Egypt, Ghana and Ivory Coast, unidirectional but reserved causality for Cameroon, Morocco, and Nigeria and neutral relationship for Benin, Republic of Congo, Kenya, Senegal, South Africa, Sudan, Togo, Tunisia and Zimbabwe. Altinay and Karagol (2005) again used the standard Granger causality test and Dolado-Lukepohl test to examine the causal relationship between electricity consumption and economic growth for Turkey using annual data covering the period 1950-2000. Zivot and Andrews’s unit roots test with endogenous structure breaks employed in the study indicated that both series were stationary with different structural breaks. The study used both Toda Yamamoto and the Dolado- Lukepohl causality and the results from both tests suggested strong evidence of unidirectional causality running from electricity consumption to GDP. This implies that the supply of electricity is vitally important to meet the growing electricity consumption in Turkey. In another study, Wolde-Rufael (2006) reinvestigated the long-run and causal relationship between electricity consumption per capita and real gross domestic product (GDP) per capita for 17 African countries for the period 1971– 2001 using ARDL cointegration test proposed by Pesaran et al. (2001) and a Granger causality test due to Toda and Yamamoto. The empirical evidence shows that there exists a long-run relationship between electricity consumption per capita and real GDP per capita for only 9 countries and Granger causality for only 12 countries. For 6 countries, there was a positive unidirectional causality running 50    from real GDP per capita to electricity consumption per capita; an opposite causality for 3 countries and bi-directional causality for the remaining 3 countries. Squalli and Wilson (2006) investigated the electricity consumption and income hypothesis for Bahrain, Kuwait, Osman, Qatar, South Korea and United Arad Emirates. The paper used the bounds test procedure suggested by Pesaran et al (2001) and Toda and Yamanto non-causality approach to test for the long-run relationship. The data used in this study were yearly and it ranges from 1980- 2003. This paper finds evidence of a long-run relationship between electricity consumption and economic growth for all the countries. It also finds support for the efficacy of energy conservation measures in 5 of the 6 countries except Qatar. Ciarreta and Zarraga (2007) investigated linear and nonlinear causality between electricity consumption and economic growth in Spain for the period 1971 to 2005. Applying both standard Granger causality and Dolado-Lukepohl methodology, they found evidence of unidirectional linear causality running from GDP to electricity consumption. By contrast, the study found no evidence of nonlinear Granger causality between the series in either direction. Zachariadis (2007) applied different Granger causality test methods such as VEC, ARDL and Toda-Yamamoto to test causal relationship between GDP and total energy use as well as sectoral energy consumption for G-7 countries namely Canada, France, Germany, Italy, Japan, the United Kingdom and the United States, using aggregate and sectoral data over different time periods. The results of these three models were in agreement for the US but inferred large disparity for all other countries. 51    Combo (2008) also computed the direction and the dynamic interrelationship between electricity infrastructure and real GDP in the Mexican economy during 1937-2007. To test the direction and dynamics of such causality, a battery of parametric tests and non-parametric methodologies were applied. To test causality, the study used Granger causality test based in an error correction model, as well as on the methodology proposed by Dolado-Lukepohl of an unrestricted VAR model. The results suggested that economic growth in Mexico preceded the expansion of the electricity infrastructure in most of the sample. Furthermore, Chandran, Sharma and Madhavan (2009) also reinvestigated the link between electricity consumption and real GDP for the case of Malaysia. They have found cointegration among electricity consumption, consumer prices and real GDP by employing ARDL bounds testing. The results of the autoregressive distributed lag (ARDL) estimates of long-run elasticity of electricity consumption on GDP are found to be around 0.7 and statistically significant. Finally, in the short-run, the results of the causality test showed that there is a unidirectional causality which flows from electricity consumption to economic growth in Malaysia. Also, the relationship between electricity consumption and economic growth is investigated by Abosedra, Dah and Ghosh (2009) using time series but monthly data covering the period January 1995 to December 2005 for the case of Lebanon. Empirical results of the study confirm the absence of a long-term equilibrium relationship between electricity consumption and economic growth in Lebanon but there exist unidirectional causality running from electricity 52    consumption to economic growth when examined in a bivariate vector autoregression framework with changes in temperature and relative humidity as exogenous variables. Adom (2011) also examined the causal relationship between electricity consumption and economic growth using a methodology based on the Toda- Yamamoto test for causality relationship and the bounds testing (ARDL). Using time-series data for the period 1971- 2008, the study found is unidirectional causality that runs from economic growth to electricity consumption for Ghana. The study also found a long-run relationship between electricity consumption and economic growth. This result is also consistent with the result obtain by Wolde- Rufael (2006) but contradict the result by Akinlo (2008). Shahbaz and Lean (2011) also investigated the relationship among energy consumption, financial development, economic growth, industrialization and urbanization in Tunisia from 1971-2008. The autoregressive distributed lag bounds testing approach to cointegration and Granger causality tests were employed for the analysis of the study. The result confirmed the existence of long-run relationship between energy consumption, economic growth, financial development, industrialization and urbanization in Tunisia. Moreover, financial development, industrialization and urbanization were found to be positively related to energy consumption especially in the long-run. Long-run bidirectional causal relationships between financial development and energy consumption, financial development and industrialization, and industrialization and energy consumption were also revealed by the study. 53    Similarly, Ozturk and Acaravci (2011) using an ARDL Bounds cointegration approach investigated the relationship and the direction of causality between electricity consumption and economic growth for 11 Middle East and North Africa countries (MENA) from 1990-2006. The unit roots tests results indicated that some of the variables for Algeria, Jordan, Tunisia and United Arab Emirates do not satisfy the underlying assumptions of the ARDL bounds test approach of cointegration methodology. Therefore, before proceeding to the estimation stage, the study drops these countries from the ARDL bounds test approach of cointegration and the causality text analysis. The authors found no unique evidence of long-run equilibrium relationship between electricity consumption and economic growth in Iran, Morocco and Syria, hence, were eliminated from the sample. However, the study found the existence of level relationship between electricity consumption and economic growth for Egypt, Israel, Oman, and Saudi Arabia. The test of causality revealed a one way short- run Granger causality from economic growth to electricity consumption in Israel. In Egypt, Oman, and Saudi Arabia, the causality test revealed the existence of one-way both short and long-run Granger causality from electricity consumption to economic growth. Shahbaz, Tang and Shabbir (2011) also examined the relationship between electricity consumption, economic growth, and employment in Portugal using the cointegration methodology and the study examined the presence of a long-run equilibrium relationship using the bounds testing approach to cointegration within the Unrestricted Error-Correction Model (UECM). Using the ARDL bounds test, 54    the results shows that electricity consumption, economic growth, and employment in Portugal are cointegrated and there is bidirectional Granger causality between the three variables in the long-run. With the exception of the Granger causality between electricity consumption and economic growth, the rest of the variables also bidirectionally Granger causes each other in the short-run. Furthermore, the study found that there is unidirectional Granger causality running from electricity consumption to economic growth, but no evidence of reversal causality in the short run is found. Lee (2005) examined the causality issue between energy consumption and GDP for 18 developing countries over the period 1975 to 2001 using three different panal unit roots, heterogeneous panel cointegration and panel based error correction model. Empirical evidence based on the trivariate model of energy consumption, GDP and capital indicated that there exists a long run equilibrium relationship among these variables after adding specific heterogeneous country effects and energy consumption Granger causes GDP in both the short run and the long run. This result indicates that energy conservation may harm economic growth in developing countries regardless of being transitory or permanent. Lee and Chang (2007) in a related study employed data on 22 developed countries from 1965 to 2002 and 18 developing countries from 1971 to 2002 to test the causal relationship between energy consumption and GDP growth. They used the panel stationary test with multiple structural breaks proposed by Carrion- i-Silvestre et al. (2005) and found both energy consumption and GDP series to be stationary series for both developed and developing countries. Applying panel 55    VAR, the study revealed a bidirectional causality causal relationship between energy consumption and economic growth for developed countries and unidirectional causal relationship running from economic growth to energy consumption for developing countries. Finally, from the impulse response functions, all of the variables in the panel VAR had a positive effect on each other, but their impact was greater and more persistent in developing countries. Their result for developing countries is inconsistent with the result of an earlier work done by Lee (2005). In another study, Nondo and Kahsa (2009) investigated the long run relationship between energy consumption and GDP for a Panel of 19 African countries based on annual data for the period 1980-2005. The results showed that GDP and energy consumption are integrated of order one and they move together in the long run indicating that they are cointegrated. Their study estimated the long run relationship and the test of causality using panal-based error correction model. The results indicated that in the short run, the neutral hypothesis holds however in the long run causality is unidirectional, running from energy consumption to GDP. Nguyen-Van (2010) also used the semi parametric partially linear panel model and Westerlund and Edgerton test, which allows for structural breaks in the panel to examine the relationship between energy consumption and income for the panel of 158 countries for the period 1980-2004. The results suggested that energy consumption increases with income but the effect of changes in energy 56    consumption structure, specifically, the natural gas and petroleum share, is not significant. Acaravci and Ozturk (2010) applied panel data to investigates the long- run relationship and causality issues between electricity consumption per capita and economic growth per capita in 15 transition countries (Albania, Belarus, Bulgaria, Czech Republic, Estonia, Latvia, Lithuania, Macedonia, Moldova, Poland, Romania, Russian Federation, Serbia, Slovak Republic and Ukraine) using the Pedroni panel cointegration test for the 1990–2006 period. The Pedroni panel cointegration tests employed by the study confirm no cointegration relationship electricity consumption per capita and economic growth per capita and hence the error-correction mechanisms and the causality tests could not be run for further steps in the long-term to investigate the causality between electricity consumption and economic growth. Yoo and Kwak (2010) also applied the panel cointegration test to investigate the causal relationship between electricity consumption and economic growth among seven South American countries, namely Argentina, Brazil, Chile, Columbia, Ecuador, Peru, and Venezuela for the period 1975–2006. The results indicated that the causal nexus between electricity consumption and economic growth varies across countries. There is a unidirectional, short-run causality from electricity consumption to real GDP for Argentina, Brazil, Chile, Columbia, and Ecuador. This means that an increase in electricity consumption directly affects economic growth in those countries. In Venezuela, there is bidirectional causality between electricity consumption and economic growth. This implies that 57    an increase in electricity consumption directly affects economic growth and that economic growth also stimulates further electricity consumption in that country. However, no causal relationships exist in Peru. Apergis and Payne (2011a) examined the relationship between renewable and non-renewable electricity consumption and economic growth for 16 emerging market economies within a multivariate panel framework over the period 1990– 2007. The Pedroni and heterogeneous panel cointegration tests employed in the study indicates that there is a long-run equilibrium relationship between real GDP, renewable electricity consumption, non-renewable electricity consumption, real gross fixed capital formation, and the labour force. However, the long-run elasticity estimate for renewable electricity consumption is positive, but statistically insignificant. The results from the panel error correction model revealed unidirectional causality from economic growth to renewable electricity consumption in the short-run and bidirectional causality in the long-run. Furthermore, there is bidirectional causality between non-renewable electricity consumption and economic growth in both the short and the long-run. In another large panel model of 88 countries over the 1990~2006 period, Apergis and Payne (2011b) classifies income level into 4 panels (high, middle high, middle low, and low incomes) and employed a panel cointegration technique to discover cointegration relations among electricity consumption, real gross fixed capital formation, GDP and labour. The causality test of the VEC model employed by the study indicated a bidirectional causality relationship between electricity consumption and economic growth for high income and 58    middle-high income panels. The results of the study also indicated that there was unidirectional causality between electricity consumption and economic growth in the short run for the middle-low panel but there exist a feedback relationship between them in the long run. Furthermore, the results of the study reveal that for low income panel, electricity consumption Granger causes economic growth. Conclusion This chapter reviewed relevant literature on the Ghanaian economy and its electricity sector development as well as theoretical and empirical work on the relationship between energy (electricity) consumption and economic growth. From the literature reviewed, it will be ambiguous to assign any positive or negative expectation on the impact of energy consumption on Economic growth in Ghana. This is due to the fact that the empirical literature on the nature and direction of causal relationships among the variables are inconclusive and mixed as a result of institutional, structural and policy differences of the countries involved, variety of variables and data span chosen as well as methodological differences. However, as shown by the review, growing body of empirical studies have demonstrated a strong positive link between energy (electricity) consumption and long-run economic growth. 59    CHAPTER THREE METHODOLOGY Introduction The aim of this chapter is to develop and specify an empirical model that captures the relationship between electricity consumption and economic growth in Ghana. Model Specification In order to examine the relationship between economic growth and electricity consumption, an endogenous growth model in the form of a Cobb- Douglas production function is formulated as given in equation (1). (1)t t tY AK LF α β= Where Y denotes the aggregate output at time t, K is the aggregate capital stock at time t, LF denotes labour force at time t while A denotes total factor productivity (TFP). α and β are the coefficients of elasticity for capital and labour respectively. The TFP captures growth in output not accounted for by increase in the physical input (capital and labour) in the model. From the literature, there are a large number of potential variables that can affect the TFP in this case. However, due to data availability and following 60    Shahbaz and Lean (2011); Bashiru (2011); Oh and lee (2004a) and Stern (1999), the study examined the following variables of interest resulting in: 3 5 62 4( , , , , ) M (2)A f EC FD INF D M EC FD INF Dβ β ββ β= = By substituting (2) into (1) and by specifying an extended Cobb-Douglas production function to represent the production technology of an economy, the study obtain; 3 5 6 71 2 4 M D (3) t t t t t tY K EC FD INF LF β β β ββ β β= Where Y is real GDP, K is capital stock, EC is electricity consumption, FD is financial development, INF is inflation, LF is Labour force, M is a dummy variable for constitutional regime and D is a dummy variable for economic reform.By taking logarithm of the variables involved in equation (3) and differencing real GDP, the study estimated a log-linear growth model of the form 0 1 2 3 4 5 6 7ln ln ln (4)t t t t t tY K EC FD INF LF M DΔ β β β β β β β β ε= + + + + + + + + Where ln denotes natural logarithm, Δ Y is the economic growth, EC, K, FD, INF, LF, M and D have already been defined. The coefficients 1β , 2β , 3β , 4β , 4β , 5β , 6β and 7β are the elasticities of the respective variables, 0β is the drift components, t denotes time and ε is the error term. A priori Expected Signs The following are the a priori expected signs: 1β >0, 2β >0, 3β >0, 4β <0, 5β >0 , 6β >0 and 7β >0. The justification of the a priori expected signs and the measurement of the variable are clearly shown in the next page. 61    Justification and Measurement of Variables Measuring Economic Growth (Y) Economic growth refers to the steady growth in the productive capacity of an economy over a given period of time. Following from literature, changes in real gross domestic product is used as a measure for economic growth for this study (Wolde-Rufael, 2006; Esso, 2010; Bashiru, 2011; King & Levine, 1993). Gross domestic product (GDP) is the sum of gross value added by all resident producers in the economy plus any product taxes and minus any subsides not included in the value of the products. It is calculated without making deductions for depreciation of fabricated assets or for depletion and degradation of natural resources (World Bank, 2011). Many researchers use the GDP deflator and the consumer price index interchangeably to deflate the nominal GDP. The GDP deflator to some extent is considered to be more efficient than the CPI because the deflator considers both producer and consumer goods whereas the CPI only covers consumer goods and services. The deflator is therefore used to deflate the nominal GDP to obtain the real GDP for the study. Measuring Electricity consumption Electricity consumption measured in Kilo Watts per hour (KWh) is defined as the amount of electricity consumed by each country or region in the year specified. This variable includes electricity from all energy sources and its accounts for the amount of electricity consumed by the end user, meaning that 62    losses due to transportation, friction, heat loss and other inefficiencies are not included in this figure. This variable has been used extensively in numerous works to measure electricity consumption (Adom, 2011; Akinlo, 2008; and Wolde-Rufael, 2006). The role of electricity consumption has been widely recognized as a growth-enhancing factor in developing countries (Khan and Ahmed, 2009). The effects of electricity consumption in an economy are normally believed to augment productivity, increase in employment and also encourage access to new technologies. The study therefore expects the coefficient of electricity consumption to be positively related to economic growth (β2>0). Other variables in the model: Financial development (FD): Financial development is usually defined as a process that marks improvement in quantity, quality, and efficiency of financial intermediary services (Abu-Bader & Abu-Qarn, 2008). Generally, financial development (FD) entails increasing financial intermediation, raising pension funds, expanding bonds and equity markets and tapping international sources of capital. It is expected to stimulate economic growth by enlarging the services provided by financial intermediaries such as savings mobilisation, project evaluation, and risk management all things being equal. Following from literature, the study used M2 as a percentage of GDP as a proxy for financial development. Money supply (M2) consists of M1 which 63    comprises currency, that is, paper money and coins in the hands of the non-bank public, checkable deposits in commercial banks and other depository institutions plus savings and time deposits or quasi money. M2/GDP represents the ratio of money stock to Gross Domestic Product (GDP). It has been employed in most studies as the standard measure of financial development because it has been found to be a good one and in most cases data are readily available (King & Levine, 1993). A high M2/GDP is expected to promote economic growth, all things being equal. The study therefore expects β3>0. Macroeconomic Instability (inflation) Inflation as measured by the consumer price index reflects how much the weighted price of a basket of consumer goods has changed over a given period of time. The study used this variable as an indicator to capture macroeconomic instability. The inflation rate indicates the overall ability of the government to manage the economy. A higher inflation rate implies that the government has lost control and this may have detrimental effect on the output of the economy. Also, inflation may affect the real sector through the banking system by reducing the overall amount of credit available to businesses. Higher inflation reduces the real rate of return on assets which in turn discourages saving and subsequently impacts on economic growth negatively (Quartey, 2010). The study therefore expects this variable to be negatively related to growth (β4<0). 64    Capital Stock (K) Gross fixed capital formation (K) formerly gross domestic fixed investment includes plants, machinery and equipment. It also includes the construction of roads, railways, and others such as schools, offices, hospitals, private residential dwellings, and commercial and industrial buildings. The variable is used as a proxy for capital stock and is expected to exert a positive impact on real GDP. As mentioned in Barro, and Sala-i-Martin (1992), countries that invest more tend to grow faster than those countries that save and invest less. Some African countries like Ghana which had low growth rates accounted only for a small percent of investments in GDP. Meanwhile, in such Asian countries like the four Asian Tigers (South Korea, Singapore, Hong Kong and Taiwan) investments yielded up to 50 percent of GDP. Consequently, the study expects the coefficient of capital to be positive (β1>0) on a priori and theoretical grounds, thus the higher the rate of investment, the higher the rate of real GDP growth, all things being equal. This is in line with both the neoclassical and endogenous growth predictions. Gross fixed capital formation as a proxy for capital has also been used in several other studies such as Aryeetey and Fosu (2005). It is also worth emphasizing that the capital used in the study refers to real capital calculated using 2000 constant prices. Labour Force (LF) For the labour input in this study, labour force which constitutes the percentage of the total population aged 15 to 65 years, who are active and 65    economically productive and is expected to have a positive impact on growth is considered as an appropriate measure of labour input. The reason for choosing this variable is its authenticity in empirical literature on growth (Jayaraman and Singh, 2007) as there can be no growth achievement without the involvement of labour as a factor input. Solow (1956) and Swan (1956) also advised that labour force should be included in an endogenous growth model because of its impact on the work force and this has been proven empirically in many researches. All things being equal, the higher the labour force the higher the supply of labour and hence output. Therefore, the coefficient of labour is expected to be positive ( 5β ). Constitutional Regime Variable (M) The constitutional regime variable as a dummy variable is included to capture the impacts of constitutional rule that existed during the period under consideration. It is a binary variable which takes the value of zero for the period in which there was no constitutional regime and one for the period under constitutional rule. Thus, D=1 from 1971; 1980-1981; 1993-2010 and D=0 from 1972-1979; 1982-1992. A regime of constitutional rule ensures well functioning democratic institutions, which is a precondition for a favourable investment climate. Non-constitutional transfers of executive power (thus coups) are particularly likely to increase uncertainty (Stasavage, 2002). Thus, a socio- politically stable environment where property rights and contracts are enforced through a properly functioning judicial system will have a positive impact on 66    private investment and hence economic growth. Hence, the coefficient of this dummy variable in the model is expected to be positive (β6 >0). Economic Reform Variable (D) The economic reform variable used in this study is constructed such that it takes the value of one (1) in periods after the reforms (1983-2010) and zero (0) otherwise. The variable is included in the model to discern the effect of economic reforms on economic growth. This variable is therefore expected to have a positive impact on GDP growth. Hence, the coefficient of this dummy variable is expected to be positive (β7 >0). All variables are expressed as natural logarithms, with the exception of Financial development, Inflation and Labour force which were already in a preferred measure thus in percentages. Data Type and Source The study relies on secondary data. The data on all the variables were obtained from the World Development Indicators (WDI) online database. These data sets were crossed checked from various sources for consistency and were proved to be consistent with each order. The study considers a sample period of 40 annual observations for each variable ranging from 1971 to 2010. 67    Estimation Procedures To test the direction of causality between electricity consumption and economic growth, the study applied Granger causality test within the framework of cointegration and error-correction models. The testing procedure involves: 1. First, testing the time series properties of the data by using the Augmented Dickey–Fuller (ADF) and the Phillip-Perron (PP) tests. This was done by carrying out the unit roots test to determine whether the variables are stationary. 2. Second, the study proceeds to test for short-run and long-run relationships among the variables using the Autoregressive Distributed Lag (ARDL) approach otherwise known as the bounds testing approach to cointegration. 3. Thirdly, the stability and diagnostic test statistics of the ARDL model is examined to ensure the reliability and the goodness of fit of the model. 4. Finally, the study employed Granger-causality to test for the causality relationship between electricity consumption and economic growth. The causality test is preceded by cointegration testing since the presence of cointegrated relationships have implications for the way in which causality testing is carried out. Unit Roots Test Confirming the order of integration is a pre-requisite for almost all time series analysis. Time series data are mostly non stationary in level forms and that 68    regression involving non-stationary time series often lead to the problem of spurious regression. This occurs when the regression results reveal a high and statistically significant relationship among variables when in fact, no relationship exist. A time series is stationary if its mean, variance and autocovariances are independent of time but due to the data generating process, time series data is rarely stationary. The study used both the Phillips-Perron (PP) and the Augmented Dickey- Fuller (ADF) statistic to test for the presence of unit roots tests. This was done to ensure reliable results of the test for stationarity due to the inherent individual weaknesses of the techniques. These tests are similar except that they differ with respect to the way they correct for autocorrelation in the residuals and also the ADF test is low power in small sample (Cheung and Lai, 1993), so the study applied the PP unit roots tests to check the robustness of the estimation results. The PP nonparametric test generalises the ADF procedure, allowing for less restrictive assumptions for the time series in question The null hypothesis to be tested is that the variable under investigation has a unit roots against the stationarity alternative. In each case, the lag-length is chosen using the Akaike Information Criterion (AIC) and Schwarz Information Criterion (SIC) for both the ADF and PP test. Both the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) or Schwarz Information Criterion (SIC) have the common objective of selecting a model that produces errors that approach a white noise process as much as possible, subject 69    to the constraint that the smallest possible number of lag terms or estimated parameters is included to ensure parsimony as well (Bashiru, 2011). The sensitivity of ADF tests to lag selection renders the PP test an important additional tool for making inferences about unit roots. The basic formulation of the ADF is specified as follows: 1 1 1 1 (5) P t t i t t i X t X Xα δ ρ λ ε− − = Δ = + + + Δ +∑ Where Xt denotes the series at time t, Δ is the first difference operator,α ,δ , ρ , λ are parameters to be estimated and ε is the stochastic random disturbance term. It is widely known that the ADF tests do not consider cases of heteroscedasticity and non-normality that are regularly disclosed in raw data of economic time series variables, and are also unable to discriminate between stationary and non stationary series that have a high degree of autocorrelation. The PP test for unit roots is therefore employed in the empirical analysis in order to resolve this problem. The PP test is also superior to the ADF test in situations where the time series variables under consideration have serial correlation and a structural break. This is based on the assumptions inherent in both tests. The ADF test assumes the error terms are independent with a constant variance whereas the PP test assumes the error terms are weakly dependent and heterogeneously distributed and thus provides robust estimates over the ADF and is specified as follows: ( )2 1 2 1 /2 (6) m t t i t i t i X X t Tα λ θ θ ε− − = Δ = + + − + Δ +∑ 70    In both equations (5) and (6), 1 2,i iε ε are the covariance stationary random error terms. The following hypotheses are therefore tested in both situations: 0 1 : 0 : 0 H H ρ ρ = ≠ The null hypothesis is that the series contains unit roots, implying non stationary against the alternative hypothesis that it does not contain unit roots, implying stationary. The decision rule is that, if the ADF and PP statistics are higher (in absolute terms) than the critical values, we fail to accept the null hypothesis and conclude that there is no unit root implying stationary. Also, if the ADF and PP statistics are less negative than the critical values then we fail to reject the null hypothesis and conclude that there is unit root implying non stationary. Autoregressive Distributed Lag (Bounds Test) Approach to Cointegration A large number of past studies have used the Johansen cointegration technique to determine the long-term relationships between variables of interest. In fact, this remains the technique of choice for many researchers who argue that this is the most accurate method to apply for I(1) variables. Recently, however, a series of studies by Pesaran and Shin (1999), Pesaran and Pesaran (1997) and Pesaran et al. (2001) have introduced an alternative cointegration technique known as the ‘Autoregressive Distributed Lag (ARDL)’ bound test. This technique has a number of advantages over Johansen cointegration techniques. First, the ARDL model is the more statistically significant approach to determine 71    the cointegration relation in small samples, while the Johansen co-integration techniques require large data samples for validity. In other words, the ARDL approach is more robust and performs better for small sample sizes (such as in this study) than other co-integration techniques (Pesaran & Shin, 1999). A second advantage of the ARDL approach is that while other cointegration techniques require all of the regressors to be integrated of the same order; the ARDL approach can be applied whether the regressors are I(1) and/or I(0). This means that the ARDL approach avoids the pre-testing problems associated with standard cointegration, which requires that the variables be already classified into I(1) or I(0) (Pesaran et al., 2001). For instance, if we are not sure about the unit roots properties of the data, then applying the ARDL procedure is the more appropriate model for empirical work. As Bahmani- Oskooee (2004) explains, the first step in any cointegration technique is to determine the degree of integration of each variable in the model but this depends on which unit roots test one uses since different unit roots tests could lead to contradictory results. For example, applying conventional unit roots tests such as the Augmented Dickey Fuller and the Phillips-Perron tests, one may incorrectly conclude that a unit roots is present in a series that is actually stationary around a one-time structural break (Perron, 1991) The ARDL approach is useful because it avoids these problems. Another difficulty of the Johansen cointegration technique which the ARDL approach avoids concerns the large number of choices which must be made: including decisions such as the number of endogenous and exogenous 72    variables (if any) to be included, the treatment of deterministic elements, as well as the order of VAR and the optimal number of lags to be used. The estimation procedures are very sensitive to the method used to make these choices and decisions (Pesaran & Shin, 1999). Finally, with the ARDL approach it is possible that different variables have different optimal numbers of lags, while in Johansen- type models this is not permitted. According to Pesaran and Pesaran (1997), the ARDL approach requires the following two steps. In the first step, the existence of any long-term relationship among the variables of interest is determined using an F-test. The second step of the analysis is to estimate the coefficients of the long-run relationship and determine their values, followed by the estimation of the short- run elasticity of the variables with the error correction representation of the ARDL model. By applying the ECM version of ARDL, the speed of adjustment to equilibrium will be determined. In this study, following Pesaran et al. (2001) and Shahbaz and Lean (2011), the short run and long run elasticities are estimated by following the Unrestricted Error Correction Model (UECM) which has unrestricted intercepts and no trends based on the assumption made by Pesaran et al. (2001) as: 21 1 1 3 1 4 1 5 1 6 1 1 1 2 3 4 5 6 1 1 1 1 1 ln ln ln ln ln (7) p t o t t t t t t i t i i p p p p p i t i i t i i t i i t i i t i t i i i i i LnY Y K EC FD INF LF lnY K EC FD INF LF β β β β β β β α α α α α α υ − − − − − − − = − − − − − = = = = = Δ = + + + + + + + Δ + Δ + Δ + Δ + Δ + Δ + ∑ ∑ ∑ ∑ ∑ ∑ Where Δ denotes the first difference operator, ߩ is the lag order selected by Akaike’s Information Criterion (AIC), 0β is the drift parameters while tυ is 73    white noise error term which is ~N (0, δ2). The parameters ijα are the short-run parameters and ijβ are the long-run multipliers. All the variables are defined as before. The study begins by estimating equation (7) with the bounds test by applying the OLS method which is normally the first procedure in the ARDL model. The F-test or Wald test is used to test for the presence of long-run relationship among the variables in equations (7) given as follows: The null hypotheses of no long-run relationship among the variables in equations (7) is tested against the alternative hypotheses of a long- run relationship as follows: ܪ : ߚଵ ߚ ߚ ߚ ߚ ଺ ߚ଻ 0 ܪଵ: ߚଵ ≠ ߚଶ ≠ ߚଷ ≠ ߚସ ≠ ߚହ ≠ ߚ଺ ് ߚ଻ ്0 ଴ = ଶ = ଷ = ସ = ହ =ߚ = ൌ The existence of cointegration between the variables under consideration is tested based on the F-statistics or Wald statistics. Given that the asymptotic distribution of F-statistics is non-standard without considering the independent variables being I(0) or I (1), Pesaran and Pesaran (1997) have provided two sets of critical values for the different numbers of regressors (k), and whether the ARDL model contains an intercept and/or trend. Therefore, the calculated F-statistic is compared with these sets of critical values developed on the basis that the independent variables are I(d) (where 0 1d≤ ≤ ). The lower critical bound assumes that all the variables are I(0), meaning that there is no co-integration among the variables, while the upper bound assumes that all the variables are I(1). so if the calculated F- statistic falls outside the upper critical value, then a null hypothesis of no cointegration will be rejected 74    regardless of whether the variables are I (0) or I (1) implying a long- run relationship among the variables. However, if the F-statistic falls below the lower bound then the null hypothesis of no co-integration cannot be rejected. If the F-statistic lies within the lower critical and upper critical bounds, the test is inconclusive and its depends on whether the underlying variables are I(0) or I(1). This necessitates the testing for unit roots on the variables under investigation (Pesaran & Pesaran, 1997). In order to obtain the optimal lag length for each variable, the ARDL methodology estimates 1( 1)km ++ number of regressions, where p is the maximum number of lags to be used and k is the number of variables in the equation. The lag of the ARDL model is based on the Schwarz-Bayesian Criterion (SBC) or the Akaike Information Criterion (AIC) or the Hannan and Quinn (HQ) criterion. The SBC uses the smallest possible lag length and is therefore described as the parsimonious model. Given that cointegration has been established from the ARDL model, the long-run and error correction estimates of the ARDL and their asymptotic standard errors are then obtained. 31 2 4 5 0 1 2 3 4 5 1 0 0 0 0 6 1 2 0 ln ln ln ln (8) qq q qp t t i t i t i t i i i i i i q t i t i Y Y EC K FD IN LF M D μ β β β β β β λ λ ν − − − − = = = = = − = = + + + + + + + + ∑ ∑ ∑ ∑ ∑ ∑ t iF − The ARDL error correction representation of the series is also estimated as 75    0 1 2 3 4 1 0 0 0 5 6 1 1 2 0 0 ln ln ln ln (9) p p p p t i t i i t i i t i i t i i i i i p p i t i i t i t t i i Y Y EC K FD INF LF ECT M D φ δ δ δ δ δ δ γ λ λ ν − − − − = = = = − − − = = Δ = + Δ + Δ + Δ + Δ + Δ + Δ + + + + ∑ ∑ ∑ ∑ ∑ ∑ Where γ is the speed of adjustment of the parameter to long-run equilibrium following a shock to the system and 1tECT − is the residuals obtained from equations (8). The coefficient of the lagged error correction term γ is expected to be negative and statistically significant to further confirm the existence of a co-integrating relationship among the variables in the model. The diagnostic test statistics of the selected ARDL model can then be examined from the short-run estimates at this stage of the estimation procedure to ensure the reliability of the goodness of fit of the model. The test for parameter stability can also be performed at this stage by plotting the cumulative sum of recursive residuals (CUSUM) and the cumulative sum of squares of recursive residuals (CUSUMSQ) to know whether the coefficients of the estimated model are stable over the study period. Granger Causality Test The study of causal relationships among economic variables has been one of the main objectives of empirical econometrics. According to Engle and Granger (1987), cointegrated variables must have an error correction representation. One of the implications of Granger representation theorem is that if non-stationary series are cointegrated, then one of the series must Granger cause the other. This statement of causality means that tX causes if the past tY 76    history of tX can be used to predict more accurately than simply using the past history of only. Y is said to be Granger-caused by X if X helps in the prediction of Y, or equivalently if the coefficients of the lagged X’s are statistically significant. It is important to note that the statement ‘X Granger causes Y’ does not imply that Y is the impact or the result of X. tY tμ η tY Granger causality measures precedence and information content but does not by itself indicate causality in the more common use of the term. This type of is causality is importantly for at least two reasons. First, it is equivalent to the econometric exogeneity in the sense that unidirectional causality that runs from the explanatory variables to the dependent variables serves a prerequisite for the consistent estimation of distributed lag models that do not involve lagged dependent variables. Second, it can be likened to leading indicators and rational expectations. To examine the direction of causality in the presence of cointegrating vectors, Granger causality is conducted based on estimating: 1 1 (9)t i t i i t i i i X Y Xα β ∞ ∞ − − = = = + +∑ ∑ 1 1 (10)t i t i i t i t i i Y X Yλ λ ∞ ∞ − − = = = + +∑ ∑ Where ( 1, 2,..., )i iα = = ∞ so that fails to causetY tX . The error terms are assumed to fulfill the criteria E( tμ )= E( tη )= E( t sμ μ )= E( t sμ μ )=0; and E( t tμ μ )= 2 μδ , E( t t ) = 2ηδ . ηη 77    Data Analysis The study employed both descriptive and quantitative analysis. Charts such as graphs and tables were employed to aid in the descriptive analysis. Unit roots tests were carried out on all variables under the study to ascertain their order of integration. Furthermore, the study adopted the Autoregressive Distributed Lag (ARDL) econometric methodology for co-integration to obtain both the short and long run estimates of the main variables involved. All estimations were carried out using Microfit (4.1) and Econometric views (Eviews) 5.0 package. Conclusion This chapter formulated the econometric model to be estimated for the study and specified the technique to be used for estimation. The methodology of this study was developed from an endogenous growth model in which electricity consumption, real GDP per capita, financial development, capital stock, inflation and labour force are the main variables of interest. Explanations of the various tests of stationarity that was adopted for this study were also shown. The study further described the sources and data set used in carrying out the study and also described the variables used in the research. The ARDL approach to co-integration and error correction models are used to determine the adjustment to equilibrium among the key variables. The study finally formulates the Granger causality equations to determine the direction of causality of the variables under investigation. 78    CHAPTER FOUR RESULTS AND DISCUSSION Introduction The results of the study are presented and discussed in this chapter. The results of the descriptive statistics of the relevant variables, both ADF and PP unit roots tests, Bounds test approach to cointegration, Granger-causality test, and stability and diagnostics tests are presented and discussed. These results are discussed in relation to the hypotheses of the study. Descriptive Statistics In this section, an analysis of the descriptive statistics is carried out. Table 1 illustrates these statistics. From Table 1, the total number of observations used was 40 and it was found that all the variables have positive means. Further examination of the table reveals that all the variables are slightly positive skewed with the exception of electricity consumption meaning that extreme changes of the series were not recorded within the observed period. The deviation of the variables from their means as shown by the standard deviation gives indication of wide growth rate (fluctuation) of these variables over the study period. 79    The Jarque-Bera statistic also shows that the null hypothesis that the series are drawn from a normally distributed random process can be rejected for all the variables since their probability values are not statistically significant with the exception of electricity consumption, capital stock and inflation. The probability values of these variables (electricity consumption, capital stock and inflation) are significant as shown in Table 1 below. Table 1: Summary Statistics LY LEC LK FD INF LF Mean 6.54442 22.2220 2.93852 19.3763 33.1699 53.6532 Median 6.52206 22.2605 2.99353 19.5533 28.1735 53.0920 Maximum 6.89936 22.6587 3.88778 29.4938 123.061 57.6005 Minimum 6.25225 20.8639 1.20546 9.68406 5.18907 51.1832 Std. Dev. 0.15802 0.37277 0.62683 5.55543 22.9398 2.22448 Skewness 0.31840 -1.5857 0.82271 0.06186 2.00282 0.43951 Kurtosis 2.48593 6.10375 3.35118 2.18352 7.55684 1.71230 Jarque-Bera 1.11631 32.8202 4.71790 1.13659 61.3499 4.05138 Probability 0.57227 0.00000 0.09452 0.56649 0.00000 0.13190 Sum 261.777 888.881 117.541 775.053 1326.80 2146.13 Sum Sq. Dev. 0.9738 5.41927 15.3239 1203.65 20523.1 192.985 Observations 40 40 40 40 40 40 Note: Std. Dev. represents Standard Deviation while Sum Sq. Dev. represents Sum of Squared Deviation Source: Computed by the author using Eviews 5.0 Package 80    Stationary Test Before applying the ARDL or bounds test approach to cointegration and Granger-causality test, unit roots test was first conducted in order to investigate the stationarity properties of the data series. This was done to ensure that the variables were not integrated of order two (that is, I(2)) so as to avoid spurious results. The computed F-statistics provided by Pesaran et al. (2001) are not valid in the presence of I(2) variables. This is so because the bounds test is based on the assumption that the variables are integrated of order zero (that is, I(0)) or integrated of order one (that is, I(1)). As a result, all the variables were examined by first inspecting their trends graphically (Appendix A). From the graphs in Appendix A and B, it is clear that all the variables except inflation rate appear to exhibit behaviours of non-stationary series. However, the first differences of the variables show no trend indicating that the series are stationary in their first difference. Additionally, the Augmented Dickey-Fuller (ADF) and Phillips and Perron (PP) tests were applied to all variables in levels and in first difference in order to formally establish their order of integration. The optimal number of lags included in the test was based on automatic selection by Schwarz-Bayesian Criterion (SBC) and Akaike Information Criterion (AIC). The study presented and used the P-values for making the unit roots decision which arrived at similar conclusion with the critical values. In order to be sure of the order of integration of the variables, the test was conducted first with intercept and no time trend and second with intercept and time trend. The results of both tests for unit roots for all 81    the variables at their levels with intercept and trend and their first difference are presented in the Appendix. Appendices C, D, E and F show that at levels, the null hypothesis of the presence of unit roots for all the variables with the exception of inflation rate cannot be rejected since the P-values of the ADF and the PP statistic are not significant at all the conventional levels of significance. However, at first difference, all the variables are stationary since the unit roots hypothesis could be rejected for all the variables. The results show that real GDP per capita, electricity consumption measured in kilo watts per hour, capital stock, labour force and financial development are each integrated of order one or I (1) while inflation is integrated of order zero or I (0) according to both the ADF and PP statistics. Therefore, in order to eliminate the possibility of spurious regression results, the first difference of the variables was employed in the estimation process. Cointegration Analysis Johansen (1991) argued that cointegration can be used to establish whether there exists a linear long-term economic relationship among the variables of interest. Therefore, since the aim of this study is to examine the relationship between electricity consumption and economic growth, it is essential to test for the existence of long-run equilibrium relationships between the variables within the framework of the bounds testing approach to cointegration. Given a small sample size and the use of annual data, a lag length of 2 is used in the bounds test. 82    Pesaran and Shin (1996) suggested a maximum lag length of 2 for annual data in the bounds test approach. The initial step of the ARDL approach is to estimate the conditional UECM model by ordinary least square in order to test for the presence of long run relationship among the variables. This is done by conducting an F-test for the joint significance of the coefficients of the lagged level variables in the model. Thus, each of the variables in the model is taken as a dependent variable and a regression is run on the others. For instance, real GDP per capital is taken as the dependent variable and it is regressed on the other variables. After that another variable for example, electricity consumption in the model is also taken as the dependent variable and it is regressed on the other variables. This action is repeated for all the variables in the model and when this is done the number of estimated regressions would be equal to the number of variables in the model. According to Pesaran et al (1997), these OLS regressions in the first differences are of no direct interest to the bounds cointegration test. It is however, the F-statistic values of all the regressions when each of the variables is normalized on the others which are of great importance. This F-statistic tests the joint null hypothesis that the coefficients of the lagged levels are zero thus indicating whether there is the existence or otherwise cointegration among the variables in the long run. Table 2 reports the results of the computed F-statistic when each variable is normalized (that is, considered as a dependent variable) in the ARDL–OLS regressions. 83    Table 2: Bounds Test for the existence of Cointegration 90% level 95% level 99% level K 1(0) 1(1) 1(0) 1(1) 1(0) 1(1) 5 2.262 3.367 2.649 3.805 3.516 4.781 Dependent Variable F- Statistic FLY(LY |LEC,FD,LK,INF,L) FLEC(LEC |LY,FD,LK,INF,LF) FFD(FD |LY,LEC,LK,INF,LF) FLK(LK |LY,LEC,FD,INF,LF) FINF(INF |LY,LEC,LK,FD,LF) FLF(L |LY,LEC,LK,FD,INF) 3.7403 1.3544 2.1454 3.4579 1.5874 2.1378 Note: Critical values are obtained from Pesaran and Persaran (1997), Appendix C, Table F, pp. 478 and K is the number of regressors in equations (7). Source: Computed by the author using Microfit 4.1 From Table 2, the calculated F-statistic reported shows that, there exists a long-run relationship between economic growth (Y) and its explanatory variables because the F-statistic value of 3.7403 is higher than the top critical bound value of 3.367 at 10 percent significance level. This implies that the null hypothesis of no cointegration among the variables in equation (7) is rejected meaning that there exists a long run relationship between economic growth and its determinant in equation (7). Again Table 2 show that when the other variables are taken as dependent variable, the calculated F- statistic values are less than the lower bound critical values, thus accepting the null hypothesis of no cointegration. However, capital 84    stock is cointegrated when used as dependent variable at 10 percent significant levels. This is because the calculated F- statistic value of FLK (.) = 3.4579 exceed the upper critical value of 3.367 at 90% significance level as shown in Table 2. These results mean that the null hypothesis of no cointegration cannot be rejected when electricity consumption, financial development, inflation and labour force are used as dependent variable. This is because their calculated F- statistic values of 1.3544, 2.1454, 1.5874 and 2.1378 respectively are less than the lower critical bound value of 2.262 at 10 percent level of significance. Since ARDL framework is a single equation model and this study is based on growth theory, economic growth variable (LY) is used as the dependent and consequently the result of the other regression that is cointegrated is neglected. Given that there is existence of cointegration among the variables in the growth equation, the study therefore proceeds with the growth equation to estimate their long run coefficients and the short-run dynamic relationships using the ARDL cointegration framework. Results of the Long-Run Relationship Given the results of the cointegration analysis, the long run relationships among the variables were estimated using the ARDL framework and the results are presented in Table 3. The Schwarz-Bayesian Criterion (SBC) and a lag length of two was used in the estimation of the ARDL model, given the annual nature of the data set. 85    As shown in Table 3, the results indicate theoretically correct and a prior expected signs for all of the explanatory variables. Electricity consumption measured in kilo watts per hour (KWh), financial development, capital stock, inflation, the dummy variable for constitutional regime and labour force all have the expected sign and exert statistically significant effect on real GDP (Y) in the long-run. The constant term is also positive and statistically significant Table 3: Long-run estimates based on ARDL Approach Variable Coefficient Standard error T-Ratio [Prob.] LEC 0.11372 0.030283 3.7553 [0.001] *** FD 0.0088897 0.0037047 2.3996 [0.024] ** LK 0.14987 0.049333 3.0379 [0.005] *** INF -0.0026229 0.0008028 -3.2673 [0.003] *** LF 1.1483 0.22450 5.1150 [0.000] *** M 0.073902 0.027269 2.7101 [0.012] ** D 0.32404 0.87267 0.3713 [0.713] C 11.4785 5.6615 2.0274 [0.052] * Note: ***,**, and * denote significance at 1%, 5% and 10% respectively Source: Computed by the author using Microfit 4.1 Holding the influence of all the variables in the model constant, the positive and statistically significant constant term in Table 3 implies that the Ghanaian economy will grow by approximately 11.5 percent due to the influence of all other variables that are not included in the model. 86    From Table 3, the long-run relationship results confirm the a priori expectation that electricity consumption contributes positively to growth of GDP since the coefficient of the electricity consumption in the long run growth equation is positive and statistically significant at 1 percent level. The coefficient of 0.11372 indicates that a 1 percent change in electricity consumption results in approximately 0.114 percent change in real GDP. This means that in the long-run, a 1 percent increase in electricity demand has the potential of stimulating economic growth in Ghana by about 0.114 percent. This result is in line with what Khan and Ahmed (2007) asserted for developing countries that increases in electricity consumption influences economic growth by augmenting productivity, increasing employment and also encouraging access to new technologies. The positive relationship between electricity consumption and economic growth concurs with the findings of Odhiambo, (2009) for Tanzania and Belloumi (2009) for Tunisia who found a positive long run relationship between electricity consumption and economic growth. Consistent with expectation, the coefficient of financial development is positive and statistically significant at 5 percent. The result specifically shows that, in the long run, if the country deepens the financial sector by 1 percent, there will be a 0.0088897 percent increase in real GDP. The positive and statistically significant effect of the financial development is consistent with the predictions of the endogenous growth theorists as well as the supply leading view of the relationship between financial development and economic growth. 87    Both the McKinnon-Shaw and the Endogenous growth theorists predict that financial deepening affects growth through investment. The endogenous growth theory emphasizes the role of the financial sector in promoting innovations, income distribution and the speed of technological progress, thus contributing to long-term economic growth (King & Levine, 1993). Consistent with the endogenous growth theory, the financial sector promotes long-run economic growth through two major channels: the volume of investment and the efficiency of investment. This result concurs with the findings of Quartey and Prah (2008) in Ghana but contradict the findings by Esso (2010) and Ahmed (2008) for Sierra Leone. Ahmed (2008) found negative but significant relationship for Sierra Leone when private sector credit was used to measure financial development. The findings by Esso (2010) showed negative impact of financial development on real GDP in the long-run. The results in Table 3 also confirms the a priori expectation that gross fixed capital formation to GDP (K) contributes positively to growth of GDP since the coefficient of capital in the long run growth equation is positive and statistically significant. The coefficient of 0.14987 indicates that a 1 percent change in capital input results in a 0.150 percent change in real GDP per capita. This means that in the long-run, a 1 percent increase in investment in physical capital has the potential of stimulating economic growth in Ghana by about 0.150 percent. Its statistical significance is shown by the p- value. This result is consistent with conclusions by Aryeetey and Fosu (2005) who obtained a positive relationship between capital and real GDP for Ghana. 88    As anticipated, inflation which is used to capture macroeconomic instability is appropriately signed. That is, the coefficient of inflation is significantly negative at 5 percent significance level. Thus the results indicate that, if the general price level increases by 1 percent, economic growth will significantly fall by 0.0026229 percent. This shows that stability of a country is an important element for achieving economic growth. The significant coefficient of the inflation variable means that if LDCs are streamlining their investment regulatory framework, implementing policies that will promote macroeconomic stability, can help achieve a higher level of economic growth (Asiedu, 2006). The negative and statistically significant effect of inflation on economic growth in the long-run is consistent with the results obtain by Bittencourt (2010) for four Latin American Countries (Argentina, Bolivia, Brazil and Peru) that inflation has a negative effect on economic growth. Ahmed and Mortaza (2005) also found a statistically significant long-run negative relationship between inflation and economic growth for Bangladesh. The results however contradict the findings by Erbaykal and Okuyan (2008). The findings by Erbaykal and Okuyan (2008) showed no statistically significant long-run relationship between inflation and economic growth for Turkey but however found a negative and statistically significant short-run relationship between inflation and economic growth. The result in Table 3 also shows that the increase in the labour force is significant to the growth of the Ghanaian economy. This is consistent with the argument of Jayaraman and Singh (2007) who asserted that there can be no 89    growth achievement without the involvement of labour as a factor input. The results specifically show that, in the long run, if labour force increase by 1 percent, there will be a 1.1483 percent increase in economic growth and this is statistically significant at 1 percent significant level. Also, consistent with expectation, the dummy variable measuring constitutional regime is positive and statistically significant at 5 percent. Specifically, the coefficient of 0.073902 indicates that a 1 percent change in capital input results in a 0.074 percent change in real GDP. This suggests that sustained improvement in economic activities is a stimulant for private investment and consequently economic growth. This finding corroborates those found by Stasavage (2002) where major constitutional change was found to stimulate the flow of private investment and consequently economic growth Finally, even though the coefficient of the dummy variable for economic reform is positive in the long run as expected, it is not statistically significant. The positive coefficient is an indication that further reform in the Ghana is likely to promote economic growth in Ghana. The error correction model that calculates the error correction term for the adjustment to short run equilibrium when there is any disequilibrium in the system as a result of a shock is given as: ECM = LY - 0.11372*LEC - 0.14987*LK -1.1483*LF +0.0026229*INF - 0.0088897*FD - 0.073902*M - 0.32404*D - 11.4785*C 90    Results of the Short Run Dynamic Model Once the long run relationships among the variables have been established within the ARDL framework, the study further estimates their short run relationships. According to Engle and Granger (1987), when variables are cointegrated, their dynamic relationship can be specified by an error correction representation in which an error correction term (ECT) computed from the long- run equation must be incorporated in order to capture both the short-run and long- run relationships. The error correction term indicates the speed of adjustment to restore equilibrium in the dynamic model. The ECM coefficient shows how quick the variables converge to equilibrium following a shock and it should have a statistically significant coefficient with a negative sign. According to Bannerjee, Dolado and Mestre (1998), the highly significant error correction term further confirms the existence of a long-run relationship. Table 5 shows the estimated short-run error correction model using the ARDL approach. From the results, the coefficient of the lagged error correction term ECM lagged one period (ECMt-1) is negative and highly significant at 1 percent significance level. This confirms the existence of the cointegration relationship among the variables in the model yet again. The ECM stands for the rate of adjustment to restore equilibrium in the dynamic model following a disturbance. The coefficient of the error correction term is 0.63359. This means that the deviation from the long-term growth rate in GDP is corrected by approximately 63 percent each year due to adjustment from the short-run towards the long-run. In 91    other words, the highly significant error correction term suggests that more than 63 percent of disequilibrium in the previous year is corrected in the current year. The rule of thumb is that, the larger the error correction coefficient (in absolute terms), the faster the variables equilibrate in the long-run when shocked (Acheampong, 2007). Therefore, the result shows that the speed of adjustment is relatively high in the model. Table 4: Estimated Short-Run ECM using the ARDL (1,0,0,0,1,0,1,0) Regressor Coefficient Standard Error T-Ratio [Prob] DLEC 0.072052 0.026569 2.7119 [0.011] ** DFD 0.0056324 0.0017389 3.2391 [0.003] *** DLK 0.094956 0.024931 3.8087 [0.001] *** DINF -0.0004274 0.0002230 -1.9162 [0.066] ** DLF 0.076396 0.0092368 8.2708 [0.000] *** M 0.046823 .019057 2.4570 [0.020] ** D 0.20531 0.53856 0.3812 [0.706] C 4.5174 1.6724 2.7011 [0.011] ** DECM(-1) -0.63359 0.14646 -4.3261 [0.000] *** Note: *** and ** denote significance at 1% and 5% respectively. Source: Computed by the author using Microfit 4.1 As shown in Table 4, the coefficient of the electricity consumption in the dynamic short run growth equation is positive and is statistically significant at 5 percent level of significance. This is consistent with the result of the long-run growth equation. From the results, a 1 percentage increase in electricity 92    consumption will induce economic growth by approximately 0.072 percent in the short run. This result indicates the crucial role that electricity plays in Ghana’s growth process as its coefficient is positive in the dynamic short run growth model just as in the long run model. Again, the result in Table 4 shows that financial development positively affects economic growth in the short run just as in the long run. Specifically, economic growth will increase by 0.0056324 percent, should the financial sector be deepened by 1 percent. This is also significant at 1 percent level of significance. This result presuppose the need to develop the financial sector to promote loans, encourage interest rates reduction since this will have beneficial effect on investment, which would eventually foster economic growth. Similarly, consistent with the long-run estimate, the coefficient of labour force maintained its positive sign and is statistically significant at 1 percent significant level. The results indicate that a 1 percent increase in labour force will increase economic growth by about 0.076396 in the short run. The positive effect of labour force is probably due to the international connections between Ghana and a host of countries such as England, China, Nigeria and others who imparted new skills and knowhow to the country labour force. Additionally, the coefficient of capital variable (K) also maintained the positive sign consistent with the long run results. Specifically, the results confirm the theoretical conclusion that capital contributes to growth of GDP since the coefficient of capital in the short run growth equation is positive and significant at 1 percent level. This reaffirms the significant role of capital in the growth process 93    of Ghana as its coefficient is positive both in the short run and the long run. From the results, the coefficient of capital is 0.094956 which indicates that a percent increase in capital input results in approximately 0.095 percent increase in economic growth, ceteris paribus. Again, this result is consistent with the results obtained by Aryeetey and Fosu (2005) for Ghana. The coefficient of inflation in the short run is also negative, consistent with the long run findings. The results thus suggest that if inflation goes up by 1 percent, economic growth falls by 0.0004274 percent. The results indicate how important it is to control inflation in the Ghanaian economy by putting in the appropriate policies. Its impact both in the short and long run appears to be debilitating. The negative relationship between inflation and economic growth as revealed in this study again is in line with empirical findings by Stockman (1981) who argued that individuals’ welfare falls whenever there is an increase in inflation. Furthermore, a relatively stable socio-political environment is relevant for accelerated growth in the Ghanaian economy. This is reflected in the positive impact of the dummy variable proxying the periods of constitutional regimes. From the results, economic growth will increase by 0.046823 if the stable political climate improves by 1 unit. This finding suggests that sustained political climate is a stimulant for private investment and consequently economic growth. Finally, consistent with the long run, the dummy variable measuring economic reform came out with a positive sign although it is not statistically significant. 94    Model Diagnostics and Stability Tests Hansen (1992) warned that estimated parameters of a time series data may vary over time. As a result, it is essential therefore to conduct parameter tests since model misspecification may arise as a result of unstable parameters and thus has the tendency of biasing the results. In order to check for the estimated variable in the ARDL model, the significance of the variables and other diagnostic and structural stability tests of the model are considered. Table 5 shows the results for the model Diagnostics and Goodness of Fit. Table 5: Model Diagnostics and Goodness of Fit R-Squared (R2) 0.75140 R-Bar-Squared 0.65933 S.E. of Regression 0.026477 F-stat. F(9, 28) 10.2012[.000] Mean of Dependent Var. 0.006042 S.D. of Dependent Var .045364 Residual Sum of Squares 0.018928 Equation Log-likelihood 90.5693 Akaike Info. Criterion 79.5693 Schwarz Bayesian Crit. 70.5625 DW-statistic 2.1032 Diagnostics LM Version F Version Serial Correlation χ2Auto (1) 1.5249[.217] 91977[.348] Functional Form χ2Reset (1) .18930[.664] .11014[.743] Normality χ2Norm (2) 3.0777[.215] Hetero χ2white (1) 1.8177[.178] 1.8085[.187] Note: χ2Auto (1), χ2Reset (1), χ2Norm (2) and χ2white (1) are Lagrange multiplier statistics for test of serial correlation, functional form misspecification, non- normal errors and heteroskedasticity, respectively. Sources: Computed by the author using Microfit 4.1 95    The diagnostic test in Table 5 shows that there is no evidence of autocorrelation and the test proved that the error is normally distributed. Additionally, the model passes the white test for heteroskedasticity as well as the RESET test for correct specification of the model. A DW-statistic of 2.1032 indicates that there is no strong serial correlation in the residuals. The overall regression is also significant at 1 percent as can be seen from the R-squared and the F-statistic in Table 5 above. The R-squared value of 0.75140 indicates that about 75 percent of the change in the dependent variable (LY) is explained by changes in the independent variables. Also, an F-statistic value of 10.2012 suggests the joint significance of the determinants in the ECM. Finally, Pesaran and Pesaran (1997) advise that we employ the CUSUM and CUSUMSQ tests when analyzing the stability of the long-run coefficients together with the short-run dynamics. The constancy of the regression coefficients is evaluated by this stability tests and they can show whether or not the regression equation is constant over time. This stability test is suitable in time series data, especially when we are uncertain about when structural change might have taken place. The null hypothesis is that the coefficient vector is the same in every period and the alternative is simply that it is not (Bahmani-Oskooee, 2004). CUSUM and CUSUMQ statistics are plotted against the critical bound of 5% significance. According to Bahmani-Oskooee (2004), if the plot of these statistics remains within the critical bound of the 5% significance level, the null hypothesis that all coefficients in the model are stable cannot be rejected. The plots of the 96    cumulative sum of recursive residuals (CUSUM) and cumulative sum of squares of recursive residuals (CUSUMSQ) stability tests as depicted in Appendix G and H indicate that all the coefficients of the estimated model are stable over the study period since they are within the 5 percent critical bounds. Granger Causality Test To find out the direction of causality between economic growth and electricity consumption, the study conducts a Granger causality test and the results are presented in Table 6. Table 6: Pairwise Granger Causality Tests Null Hypothesis F- Stat Prob. Remarks LEC does not Granger Cause LY 0.23661 0.79063 Cannot reject the null LY does not Granger Cause LEC 7.95165 0.00152*** Null is rejected FD does not Granger Cause LY 1.30798 0.06211* Null is rejected LY does not Granger Cause FD 3.02656 0.28402 Cannot reject the null LK does not Granger Cause LY 3.41439 0.04490** Null is rejected LY does not Granger Cause LK 1.34503 0.27444 Cannot reject the null INF does not Granger Cause LY 2.47882 0.09932* Cannot reject the null LY does not Granger Cause INF 5.19369 0.01094** Null is rejected LF does not Granger Cause LY 10.9112 0.00023*** Null is rejected LY does not Granger Cause LF 2.77188 0.07713* Null is rejected Note: ***, **, and * denote significance at 1%, 5%, and 10% respectively Source: Computed by the author using Microfit 4.1 97    The Granger causality test results in Table 6 suggests that the null hypothesis of electricity consumption does not Granger cause real GDP is not rejected, implying electricity consumption does not Granger cause real GDP. However, the null hypothesis that real GDP does not Granger cause electricity consumption is rejected, implying real GDP Granger causes electricity consumption. This means that there exists a unidirectional causality running from real GDP to electricity consumption. This result indicates that data on Ghana supports the Growth-led–Electricity hypothesis. The unidirectional causality running from economic growth to electricity consumption as identified in this study is not in conformity with the Bounds tests results, as the bounds test result shows there is positive and statistically significant relationship between electricity consumption and economic growth. The reason behind this may probably due to differences in estimation techniques in both the Bounds tests and the Pairwise Granger Causality and also due to the small sample size of the study. However, this finding in line with the findings of Wolde-Rufael (2006); Twerefo et al (2008) and Adom (2011) but contradicts the conclusions reached by Lee (2005) and Akinlo (2008). Lee (2005) results indicate that there is unidirectional causality between electricity consumption and economic growth and it runs from electricity consumption to economic growth while Akinlo (2008) revealed bidirectional causality between electricity consumption and economic growth for Ghana in his panel study. Similarly, the Granger causality test results in Table 6 suggests that the null hypothesis of financial development does not Granger cause real GDP is 98    rejected at 10 percent significant level, implying financial development does indeed Granger cause real GDP. However, the null hypothesis that real GDP does not Granger cause financial development is not rejected, implying real GDP Granger does not causes financial development. This shows that there is unidirectional causality running from financial development to economic growth. Furthermore, it can be observed from Table 6 that the null hypothesis that labour force does not Granger causes economic growth is rejected at 1 percent significance, also the null hypothesis that economic growth Granger causes labour force is rejected at 5 percent level of significance. This indicates that there is bi- directional causality between labour force and economic growth at 1 percent and 5 percent significance level respectively. Similarly, the results from Table 6 indicate that Gross fixed capital formation Granger causes economic growth since the study fails to reject the null hypothesis of no Granger causality at the 5 percent significance level. This means that there is unidirectional causality running from Gross fixed capital formation Granger to economic growth. Finally, the pair-wise Granger Causality tests showed a bi-directional causality between inflation and economic growth at 10 and 5 percent significance level respectively Conclusion This chapter of the study looked at the empirical results of the relationship between electricity consumption and economic growth. The chapter began by 99    examining the time series properties of the data used for the estimation. The unit roots test employing both the ADF and the PP techniques basically showed that all the series had to be differenced once to achieve stationarity except for inflation that is stationary at levels. The study further conducted the test for cointegration using the Autoregressive Distributed Lagged Model (ARDL) and Granger causality test. The results disclosed a long-run cointegrating relationship between electricity consumption and economic growth in Ghana. Both the short and long run estimates reveal a positive and statistically significant effect of electricity consumption on economic growth. This indicates the crucial role that electricity plays in Ghana’s growth process as its coefficient is positive in the dynamic short run growth model just as in the long run model. The coefficient of ECM (1) is 0.63359. This means that the deviation from the long-term growth rate in GDP is corrected by 63 percent each year due to adjustment from the short run towards the long-run. The diagnostic and parameter stability tests reveal that the model passes the tests of serial correlation, functional form misspecification, non normal errors and heteroscedasticity. The overall regression is also significant at both 5% and 1% as can be seen from the R-squared and the F-statistic and the graphs of the CUSUM and CUSUMSQ indicate the absence of any instability of the coefficients because the plots of these graphs are confined within the 5 percent critical bounds of parameter stability suggesting that all the coefficients of the estimated ARDL model are stable over the study period. 100    Finally, the Granger causality test result revealed that there is a unidirectional causality running from economic growth to electricity consumption. This result confirms that data on Ghana supports the Growth-led– Electricity hypothesis. 101    CHAPTER FIVE SUMMARY, CONCLUSIONS AND RECOMMENDATIONS Introduction This chapter summarizes the entire study. It presents summary, conclusions and recommendations derived from the analysis of the data collected. The chapter also presents the limitations and suggestions for further research. Summary of the Study Understanding the causality between electricity consumption and economic growth has been a challenge to those in the fields of growth and energy economics. Several researchers have tried to formulate models and undertake empirical works that could help in deepening the understanding of the relationships therein. But the global climate crises came as a wakeup call for all those in these fields to work harder for a better understanding of this causality. The study set itself to meet three objectives. First, to examine both the long-run and short-run relationships between electricity consumption and economic growth in Ghana and most importantly to identify the causal relationship between electricity consumption and economic growth as well as the direction of causality. 102    To examine the long-run relationship and short-run dynamic parameters of the model, the Autoregressive Distributed Lagged Model (otherwise known as the bounds testing) approach to cointegration was employed. The data series –Real GDP per capita (LY), electricity consumption (LEC), financial development (FD), capital stock (K), labour force (L) and the policy variable, inflation (INF) were first subjected to ADF and PP tests in order to ascertain their stationarity properties. The ADF and PP tests revealed that all variables except inflation were integrated of order one. The study then proceeded to examine the long-run and short-run relationships between electricity consumption and economic growth. The findings of the study are therefore presented below: 1. The cointegration analysis revealed that there exist a positive and significant relationship between GDP growth and electricity consumption in the long run. The results of the long run growth equation revealed that a 1 percent increase in electricity consumption leads to approximately 0.11% increase in GDP growth. Consistent with the long-run results, the coefficient of electricity consumption in the short-run was also positive. 2. The study also found a positive and statistically significant relationship between capital stock and GDP growth both in the long run and short run. This reemphasizes the significant role that capital plays in the growth process of Ghana. 3. Furthermore, the result of the study revealed a positive and statistically significant relationship between GDP growth and labour force both in the long run and the short run. This is consistent with the argument of 103    Jayaraman and Singh (2007) who asserted that there cannot be growth achievement without the involvement of labour as a factor input. 4. In addition to the above, financial development registered a positive impact on economic growth in the short as well as the long run. This confirms the importance of deregulating the financial sector which may reduce credit rationing to firms and consequently improves economic growth through increases in investment. 5. As anticipated, the bounds tests results also revealed that inflation both in the long and short run exerted a statistically significant negative effect on economic growth. The negative and statistically significant impact of inflation on economic growth is in line with the traditional neoclassical theory as posited by Stockman. This is an indication that high inflation episode is unattractive to economic growth in Ghana. 6. Moreover, consistent with expectation, the dummy variable measuring constitutional regime came out with a positive sign and is statistically significant both in the short run and the long run. This means that a well functioning democratic system where property rights are protected, civil society, the judiciary among other state and institutions are adequately resourced and monitored could provide a vital bridge for growth in Ghana. The coefficient of the dummy variable suggests that sustained political climate is a stimulant for private investment and consequently economic growth. 104    7. The coefficient of the dummy variable measuring economic reforms was also positive as expected, but statistically insignificant indicating that further reform is necessary to ensure growth in the Ghanaian economy. 8. Again, the coefficient of the lagged error correction term is negative and statistically significant as expected at the 1 percent significance level, suggesting that it would not take a long time for the system to return to its equilibrium once it is out of equilibrium. The size of the coefficient on the error correction term (ECM) denotes that about 63.3 percent of the disequilibrium caused by previous years’ shocks converges back to the long-run equilibrium in the current year. 9. The diagnostic tests results show that the model passes the test of serial correlation, functional form misspecification, non normal errors and heteroscedasticity. The graphs of the cumulative sum of recursive residuals (CUSUM) and the cumulative sum of squares of recursive residuals (CUSUMSQ) exhibit that there exists a stable relationship between electricity consumption and economic growth over the study period. 10. Finally, investigation of the direction of causation among the variables using the Pairwise Granger causality testing procedure was utilized. The aim was to identify any causal relationship between electricity consumption and economic growth and in which direction. The result showed a unidirectional relationship between electricity consumption and 105    economic growth and it runs from economic growth to electricity consumption. Conclusions Based on the finding of the study, the following conclusions were reached: 1. From the results, the objective of investigating the relationship between electricity consumption and economic growth was accomplished. When considering short run and long impacts, electricity consumption was found to be a stimulus for economic growth of Ghana in the short and long run. These findings reveal that there are benefits from electricity consumption as reveal in the literature. 2. Secondly, in line with empirical evidence, capital stock, labour force and financial development all serve as important determinants on economic growth in Ghana. This finding implies that capital stock, labour force, and financial development are critical in achieving sustained economic growth in the Ghanaian economy. Again, the study found statistically significant and negative effects of inflation on economic growth. This means inflation is inimical to economic growth. 3. Finally, the Granger causality test results revealed a unidirectional relationship between electricity consumption and economic growth running from economic growth to electricity consumption. This implies that economic growth Granger causes electricity consumption whereas the 106    Bounds tests results from the long-run shows that the direction of causality is from electricity consumption to economic growth. Policy Recommendations The study brought to light the importance of electricity consumption, financial development, inflation and capital investment in influencing economic growth in Ghana. Taking cognisance of this, the following recommendations are proposed to help achieve a higher and sustained economic growth in Ghana. The study found strong evidence to support the fact that electricity consumption acts as engine of economic growth and that economic performance in Ghana can be enhanced through increases in electricity demand. However, the electricity sector in Ghana is operating at low capacity level. To meet increased electricity demand, planning and investment in electricity infrastructure development is essential. In respect of this, government should intensify Public- Private Partnership (PPP) investment in the electricity sector since PPP remains one of the most innovative options that can be deployed to mobilize more resources for energy infrastructure development. Again, the Electricity Company of Ghana (ECG) could also work with other stakeholders in the electricity sector to draft and implement policies so that sustainable electricity supply may be ensured. Also, the Ghanaian economy should not rely mainly on the hydro- electricity generation (thus electricity generated from water sources) because of the cyclical nature and seasonal fluctuations of water availability. The Energy Commission and the Public Utilities Regulatory Commission (PURC) should 107    therefore devote more effort towards developing other sources of energy such as thermal, solar and wind energy for the country. The study also revealed a positive relationship between financial deepening and economic growth which is an indication that financial deepening plays an important role in growth of the economy. In respect of this, government with the Bank of Ghana and other stakeholders in the financial sector should deepen and maintain a continued implementation of the financial sector reforms in order to foster a growing financial deepening so as to promote economic growth. To ensure development of the financial sector, the Bank of Ghana in conjunction with other financial institutions could revise the conventional financial institutional lending to make it more conducive for micro-borrowing. Solidarity group lending, where a group member can guarantee for one another should also be encouraged to allow the poor access to credit at low cost without requiring collateral. This will facilitate the development of financial services leading to financial deepening. In addition, government could consider creating and maintaining macroeconomic stability necessary for energy development and competition while encouraging the stakeholders in the electricity sector to enhance technological and human resource development for efficient delivery of electricity services. Overall, the more stable the economy, the better will be the prospect of huge output growth. Therefore, inflation should be kept at a moderate rate since high rate of inflation indicate high level of distortion in Ghana. 108    Finally, the study also showed a positive relationship between capital stock and GDP growth. This implies that GDP growth could be achieved in Ghana by increasing savings so as to raise adequate capital. Despite the significant role of capital in the growth process of Ghana, there is low capital formation which has resulted in shortage of capital. The problem of shortage of capital is mainly due to low savings. Thus, increasing savings could make adequate capital available to investors. To increase savings in Ghana, one way is for the government through the central bank and other stakeholders in the banking sector to institute deposit insurance schemes to safeguard depositors. This will encourage savers to put more money at the bank. This would help mobilize adequate capital which could be channeled to investors to produce more output to increase the gross domestic product of Ghana. Also, the Bank of Ghana could also work with other stakeholders in the financial sector to draft and implement policies aimed at ensuring the ready accessibility of credit for investment at moderate cost. In addition, the Bank of Ghana together with the players in the financial market must also design policies to improve the efficiency of savings mobilisation through the banking system accompanied by efficient allocation of resources to the productive sectors of the economy. Limitations of the Study This study is not free of drawbacks. The main limitation of the study had to do with the quality and limited availability of annual data on some key 109    variables such as electricity consumption and labour force used in the study. For instance an attempt to extend the data length backwards from 1971 to 1960 or further was constrained by unavailability of data. Future Direction of Research The main focus of this study has been to examine the relationship between electricity consumption on economic growth at the aggregate level. It would be more interesting if the impact of electricity consumption on economic growth can be disaggregated in at least three major sectors: (a) the agricultural sector, (b) the industrial sector and (c) the services sector. By doing this, the channels through which electricity consumption may affect economic growth could be highlighted. Secondly, a comparative study between pre-reform and post-reform scenario pertaining to the electricity sector can be done so as to draw better conclusions. During reforms, different policies are adopted so a study can be done to evaluate the impact of these policies on economic growth. Finally, since electricity consumption is such an important element for economic growth, further support or research should be devoted towards the exact mechanism by which it influences economic growth. 110    REFERENCES Abosedra, S., Dah, A., & Ghosh, S. (2009). Electricity consumption and economic growth, the case of Lebanon. Applied Energy, 86(4), 429-432. Abu-Bader, S., & Abu-Qarn, A. S. (2008). Financial development and economic growth: The Egyptian experience. Journal of Policy Modelling, 30(5), 887-898. Acaravci, A., & Ozturk, I. (2010). Electricity consumption-growth nexus: Evidence from panel data for transition countries. Energy Economics, 32(3), 604-608. Acheampong, I. K. (2007). Testing Mckinnon-Shaw thesis in the context of lm Ghana’s financial sector liberalisation episode. Journal of Management Research and Technology, 1(2), 155-183. Adom, P. K. (2011). Electricity consumption-economic growth nexus: The Ghanaian case. International Journal of Energy Economics and Policy, 1(1), 18-31. Aghion, P., & Howitt, P. (1998). A model of growth through creative destruction. Econometrica, 60(2), 323-351. Ahmed, A. D. (2008). Financial liberalisation, financial development and growth in sub- Saharan Africa’s economic reform: An empirical investigation. Studies in Economics and Finance, 27(4), 314 – 339. Ahmed, S., & Mortaza, G. (2005). Inflation and economic growth in Bangladesh. (Policy Analysis Unit Working Paper Series: WP 0604). Retrieved February 2011, from http://www.bangladesh-bank.org 111    Akarca, A. T., & Long, T. V. (1980). Energy and employment: A time series analysis of the causal relationship. Resources and Energy, 2(3), 151-162. Akinlo, A. E. (2008). Energy consumption and economic growth: Evidence from 11 Sub-Sahara African countries. Energy Economics, 30(5), 2391-2400. Akinlo, A. E., (2009). Electricity consumption and economic growth in Nigeria: Evidence from cointegration and co-feature analysis. Journal of Policy Modelling, 31(5), 681-693. Altinay, G., & Karagol, E. (2005). Electricity consumption and economic growth: Evidence from Turkey. Energy Economics, 27(2), 849-856. Altinay, G., & Karagol, E. (2004). Structural break, unit root, and the causality between energy consumption and GDP in Turkey. Energy Economics, 26(6), 985-994. Apergis, N., & Payne, J. E. (2009). The emissions, energy consumption, and growth nexus: Evidence from the commonwealth of independent states. Energy Policy, 38(1), 650-655. Apergis, N., & Payne, J. E. (2011a). Renewable and non-renewable electricity consumption–growth nexus: Evidence from emerging market economies. Applied Energy, 88(12), 5226-5230. Apergis, N., & Payne, J. E. (2011b). Renewable and non-renewable energy consumption growth nexus: Evidence from a panel error correction model. Energy Economics, 34(3), 733-738. 112    Aqeel, A., & Butt, M. S. (2001). The relationship between energy consumption and economic growth in Pakistan. Asia-Pacific Development Journal, 8(2), 101-110. Arrow, K., & Kurz, M. (1970). Public Investment, the Rate of Return, and Optimal Fiscal Policy. Johns Hopkins University Press. Aryeetey, E., & Fosu, A. K. (2005). Economic growth in Ghana: 1960-2000. Paper presented at AERC Growth Project Workshop, Cambridge. Asafu-Adjaye, J. (2000). The relationship between energy consumption, energy prices and economic growth: Time series evidence from Asian developing countries. Energy Economics, 22(6), 615-625. Asare, A. (2008). A study on electricity end use saving potentials in residential and commercial buildings in Ghana (Unpublished master’s thesis). Aalborg University, Denmark. Retrieved February 20, from http://projekter.aau.dk/projekter/files/14550385/Abraham_Thesis-_all.pdf Asiedu, E. (2006). Foreign direct investment in Africa: The role of natural resources, market size, government policy, institutions and political instability. The World Economy, 29(1), 63-77. Ayres, R. U., & Van den Bergh, J. C. J. M. (2005). A theory of economic growth with material/energy resources and dematerialization: Interaction of three growth mechanisms. Ecological Economics, 55(1), 96-118. Bahmani-Oskooee, M. (2004). Long run demand for money in Hong Kong: An application of ARDL model. International Journal of Business and Economics, 1(2), 147-155. 113    Bannerjee, A., Dolado, J., & Mestre, R. (1998). Error-correction mechanism tests for cointegration in single equation framework. Journal of Time Series Analysis, 19(3), 267-283. Barro, R. J., & Sala-i-Martin, X. (1992). Economic growth (2nd ed.). New York: The McGraw-Hill International Editions. Belloumi, M. (2009). Energy consumption and GDP in Tunisia: Cointegration and causality analysis. Energy Policy, 37(7), 2745-2753. Bittencourt, M. (2010). Inflation and economic growth in Latin America: Some panel time series evidence. (University of Pretoria Department of Economics Working Paper, Working Paper Series: 2010-11). Retrieved October 2010, from http://web.up.ac.za Bashiru, I. (2011). Financial deepening and economic growth in Ghana (Unpublished master’s thesis). University of Cape Coast, Cape Coast. Carrion-i-Silvestre, J. L., Barrio-Castro, T. D., & Lopex-Bazo, E. (2005). Breaking the panels: An application to GDP per capita. Econometrics Journal, 8(2), 159-175. Cass, D., & Koopmans, T. C. (1965). Optimum growth in an aggressive model of capital accumulation. Review of Economic Studies, 32(2), 233-240. Chandran, V. G. R., Sharma. S., & Madhavan. K. (2009). Electricity consumption–growth nexus: The case of Malaysia. Energy Policy, 38(1), 606-612. Chebbi, H. E., & Boujelbene, Y. (2008). Agricultural and non- agricultural outputs and energy consumption in Tunisia: Empirical evidence from 114    cointegration and causality. Paper presented at the Congress of the European Association of Agricultural Economists (EAAE). Ghent, Belgium. Cheng, B. S. (1995). An investigation of cointegration and causality between energy consumption and economic growth. The Journal of Energy and Development, 21(1), 73-84. Cheng, B. S. (1997). Energy consumption and economic growth in Brazil, Mexico and Venezuela: A time series analysis. Applied Economics Letters, 4(11), 671-674. Cheng, B. S. (1998). Energy consumption, employment and causality in Japan: A multivariate approach. Indian Economic Review, 33(1), 19-29. Cheng, B. S. (1999). Causality between energy consumption and economic growth in India: An application of cointegration and error-correction modeling. Indian Economic Review, 34(1), 39-49. Cheng, B. S., & Lai, W. L. (1997). An investigation of cointegration and causality between energy consumption and economic activity in Taiwan. Energy Economics, 19(4), 435-444. Cheung, Y. W., & Lai, K. S. (1993). Finite-sample sizes of Johansen’s likelihood ratio test for cointegration. Oxford Bulletin of Economics and Statistics, 55(4), 313-328. Chontanawat, J., Hunt, L. C., & Pierse, R. (2006). Causality between energy consumption and GDP: evidence from 30 OECD and 78 non-OECD countries. (Surrey Energy Economics Discussion Paper Series 113, ISSN 115    1749-8384). Retrieved September 2011, from http://www.seec.surrey. ac.uk/research/SEEDS/SEEDS113 Ciarreta, A., & Zarraga, A. (2008). Electricity consumption and economic growth in Spain. Applied Economics Letters, 17(14), 1417–1421. Combo, E. M. (2008). Lag structure between electricity infrastructure and economic growth in Mexico. Retrieved September 2011, from http://www.webmeets.com/files/papers/LACEeALAMES/2008/855/Paper _EMCH_LAMES_Sept2011.pdf Ebohon, O. J. (1996). Energy growth and causality in developing countries. Energy Policy, 24(5), 447-453. Eggoh J. C., Rault, C., & Bangaké C. (2011). Energy consumption and economic growth: Revisited in African countries. (CESifo Working Paper No. 3590). Retrieved November 2011, from http://www.cesifo-group.de/portal/pls/ portal/docs/1/1210032.PDF. Engle, R. F., & Granger, C. J. (1987). Cointegration and error-correction representation, estimation and testing. Econometrica, 55(2), 251-278. Erbaykal, E., & Okuyan, H. A. (2008). Does inflation depress economic growth? Evidence from Turkey. International Research Journal of Finance and Economics, 13(17), 41-48. Erol, U., & Yu, E. S. H. (1987). Time series analysis of the causal relationships between energy and employment. Resources and Energy, 9(1), 75-89. 116    Esso, L. J. (2010). Long-run relationship and causality between foreign direct investment and growth: Evidence from ten African countries. International Journal of Economics and Finance, 2(2), 168-177. Fatai, K., Oxley, L., & Scrimgeour, F. (2004). Modelling the causal relationship between energy consumption and GDP in New Zealand, Australia, India, Indonesia, the Philippines and Thailand. Mathematics and Computers in Simulation, 65(4), 431–445. Fornelio, H. F. (2010). A study of the determinants of household’s demand for electricity: A case study of NV EBS (Unpublished master’s thesis). FHR Lim A Po Institute and the Maastricht School of Management, Maastricht, the Netherlands. Retrieved March 8, from http://www.fhrinstitute.org/ mod/data/viewphp?id Ghana Grid Company Limited. (2010). Ghana power reliability report. Retrieved September 2011, from http://www.gridcogh.com/site/downloads/GridCo ReportFinal.pdf Ghosh, S. (2002). Electricity consumption and economic growth in India. Energy Policy, 30(2), 125–129. Glasure, Y. U. (2002). Energy and national income in Korea: Further evidence on the role of omitted variables. Energy Economics, 24(4), 355-365. Glasure, Y. U., & Lee, A. R. (1997). Cointegration, error-correction, and the relationship between GDP and energy: The case of South Korea and Singapore. Resource Energy Economics, 20(1), 17-25. 117    Granger, C. W. J. (1986). Development in the study of cointegrated economic variables. Oxford Bulletin of Economics and Statistics, 48(3), 213-228. Granger, C. W. J. (1988). Some recent developments in a concept of causality. Journal of Econometrics, 39(2), 199-211. Guttormsen, A. G. (2004). Causality between energy consumption and economic growth. Paper presented at the Department of Economics and Resource Management, Agriculture University of Norway, Norway. Hansen, B. E. (1992). Tests for parameter stability in regressions with I(1) Processes. Journal of Business and Economic Statistics, 10(3), 321-335. Ighodaro, C. A. U. (2010). Cointegration and causality relationship between energy consumption and economic growth: Further empirical evidence for Nigeria. Journal of Business Economics and Management, 11(1), 97–111. Institute for Statistical, Social and Economic Research. (2005). Guide to electric power in Ghana. University of Ghana, Legon-Ghana. Institute for Statistical, Social and Economic Research. (2000-2009). State of the Ghanaian economy. University of Ghana, Legon-Accra Jamil, F., & Ahmad, E. (2010). The relationship electricity consumption, electricity prices and GDP in Pakistan. Journal of Energy Policy, 38(10), 6016- 6025. Jayaraman, T. K., & Singh, B. (2007). Foreign direct investment and employment creation in pacific island countries: An empirical study of Fiji. (ARTNET Working Paper No. 35/07.) Asia-Pacific Research and Training Network on Trade. Retrieved September 25, 2010, from http://www.unescap.org 118    Johansen, S. (1988). Statistical analysis of cointegration vectors. Journal of Economic Dynamics and Control, 22(3), 231-254. Johansen, S. (1991). Estimation and hypothesis testing of cointegration vectors in Gaussian vector autoregressive models. Econometrica, 59(6), 1551-1580. Jumbe, C. B. L. (2004). Cointegration and causality between electricity consumption and GDP: Empirical evidence from Malawi. Energy Economics, 26(1), 61–68. Karanfil, F. (2009). How many times again will we examine the energy–income nexus using a limited range of traditional econometric tools? Energy Policy, 37(4), 1191–1194. Kemp, M. W. (2005). The renewable energy handbook – A guide to rural energy independence, off-grid and sustainable living. Canadian, CA: Aztext Electronic Publishing Press. Khan, A., & Ahmed, U. (2009). Energy Demand in Pakistan: A disaggregate analysis. Paper presented at the 24th Annual General Meeting and Conference of the Pakistan Society of Development Economists, March 31-April 2, Islamabad. King, R. G., & Levine, R., (1993). Finance and growth: Schumpeter might be right. The Quarterly Journal of Economics, 108(3), 717-737. Kraft, J., & Kraft, A. (1978). Note and comments: On the relationship between energy and GNP. The Journal of Energy and Development, 3(2), 401-403. 119    Lee, C. C., & Chang, C. P. (2007). Energy consumption and GDP revisited: A panel analysis of developed and developing countries. Energy Economics, 29(6), 1206-1223. Lee, C. C. (2005). Energy consumption and GDP in developing countries: A cointegrated panel analysis. Energy Economics, 27(3), 415–427. Lucas, R. E. (1988). On the mechanics of economic development. Journal of Monetary Economics, 22(1), 3-42. Masih A. M. M., & Masih, R. (1998). A multivariate cointegrated modeling approach in testing temporal causality between energy consumption, real income and prices with an application to two Asian LDCs. Applied Economics, 30(10), 1287-1298. Masih, A. M. M., & Masih, R. (1996). Energy consumption, real income and temporal causality: Results from a multi-country study based on cointegration and error-correction modelling techniques. Energy Economics, 18(3), 165-183. Masih, A. M. M., & Masih, R. (1997). On the temporal causal relationship between energy consumption, real income, and prices: Some new evidence from Asian-energy dependent NICs based on a multivariate cointegration/vector error-correction approach. Journal of Policy Modeling, 19(4), 417-440. Nachane, D. M., Nadkarni, R. M., & Karnik, A. V. (1988). Co-integration and causality testing of the energy-GDP relationship: A cross-country study. Applied Economics, 20(11), 1511-1531. 120    Nguyen-Van, P. (2010). Energy consumption and income: A semiparametric panel data analysis. Energy Economics, 32(3), 557-563. Nondo, C., & Kahsai, A. (2009). Energy consumption and economic growth: Evidence from Comesa countries. Presentation at the southern Agricultural Economics Association – Annual Meeting, Atlanta, Georgia on Jan 31 – Feb 3 with number 46450. Retrieved September 2011, from http://rri.wvu.edu/wp-content/uploads/2011/07/wp2010-11.pdf. Ockwell, D. G. (2008). Energy and economic growth: Grounding our understanding in physical reality. Energy Policy, 36(12), 4600-4604. Odhiambo, N. M. (2009). Electricity consumption and economic growth in South Africa: A trivariate causality test. Energy Economics, 31(5), 635-640. Oh, W., & Lee, K. (2004a). Causal relationship between energy consumption and GDP revisited: the case of Korea 1970-1999. Energy Economics, 26(1), 51-59. Oh, W., & Lee, K. (2004b). Energy consumption and economic growth in Korea: testing the causality relation. Journal of Policy Modeling, 26(9), 973-981. Ozturk, I., & Acaravci, A. (2010). The causal relationship between energy consumption and GDP in Albania, Bulgaria, Hungary and Romania: Evidence from ARDL bound testing approach. Applied Energy, 87(6), 1938-1943. Ozturk, I., & Acaravci, A. (2011). Electricity consumption-real GDP causality nexus: Evidence from ARDL Bounds testing approach for 11 MENA countries. Applied Energy, 88(8), 2885-2892. 121    Paul, S., & Bhattacharya, R. N. (2004). Causality between energy consumption and economic growth in India: A note on conflicting results. Energy Economics, 26(6), 977–983. Pedroni, P. (1999). Critical values for cointegration tests in heterogeneous panels with multiple regressors. Oxford Bulletin of Economics and Statistics, 61(2), 653-670. Perron, P. (1991). Test consistency with varying sampling frequency. Econometric Theory, 7(03), 341-368. Pesaran, M. H., & Shin, Y. (1999). An autoregressive distributed lag modeling approach to cointegration analysis. In S. Strom, (Eds.), Econometrics and economic theory in the 20th century (Chapter 11). The Ragnar Frisch centennial symposium, Cambridge: Cambridge university press. Pesaran, M. H., & Pesaran, B. (1997). Working with microfit 4.0: Interactive econometric analysis. Oxford: Oxford University Press. Pesaran, M. H., Shin, Y., & Smith, R. J. (2001). Bounds testing approach to the analysis of level relationships. Journal of Applied Econometrics, 16(3), 289-326. Quartey, P. (2010). Price stability and the growth maximising rate of inflation in Ghana. Modern Economy, Scientific Research Paper, Legon, Ghana: ISSER, University of Ghana. (pp181-193). Retrieved March 2011, from http://www.pquartey@ug.edu.gh. Quartey, P., & Prah, F. (2008). Financial development and economic growth in Ghana: Is there a causal link? African Finance Journal, 10(1), 28–54. 122    Ramsey, F. (1928). A mathematical theory of saving. Economic Journal, 38(2), 543-559. Romer, P. M. (1986). Increasing returns and long-run growth. Journal of Political Economy, 94(5), 1002-1037. Shahbaz, M., & Lean, H. H. (2011). Does financial development increase energy consumption? The role of industrialization and urbanization in Tunisia. Energy Policy, 40, 473-479. Shahbaz, M., Tang, C. F., & Shabbir, M. S. (2011). Electricity consumption and economic growth nexus in Portugal using cointegration and causality approaches. Energy Policy, 39(6), 3529- 3536. Shiu, A., & Lam, P. L. (2004). Electricity consumption and economic growth in China. Energy Policy, 32(1), 47-54. Smulders, S., & De Nooij, M. (2003). The impact of energy conservation on technology and economic growth. Resource and Energy Economics, 25(1), 59-79. Solow, R. M. (1956). A contribution to the theory of economic growth. Quarterly Journal of Economics, 70(2), 65-94. Squalli, J., & Wilson, K. (2006). A bounds analysis of electricity consumption and economic growth in the GCC. (Working Paper -06-09), EPRU, Zayed University. Retrieved from http://www.zu.ac.ae/epru/documents/06-09- web.pdf. Stasavage, D. (2002). Private investment and political institutions. Economics and Politics, 14(1), 0954-1985. 123    Stern, D. I. (1993). Energy and economic growth in the USA, a multivariate approach. Energy Economics, 15(2), 137-150. Stern, D. I. (1999). Is energy cost an accurate indicator of natural resource quality? Ecological Economics, 31(3), 381-394. Stern, D. I. (2000). A multivariate cointegration analysis of the role of energy in the US economy. Energy Economics, 22(2), 267-283. Stern, D. I. (2003). Energy and economic growth. (Rensselaer Working Papers in Economics, 0308), Rensselaer Polytechnic Institute, USA. Retrieved February 2012, from http://www.localenergy.org//Energy%20and%20 Economic%20Growth.pdf. Stiglitz, J. E. 1974). Growth with exhaustible natural resources: efficient and optimal growth paths. Review of Economics Studies, 41(5), 123–137. Stockman, A. C. (1981). Anticipated inflation and the capital stock in a cash-in advance economy. Journal of Monetary Economics, 8(3), 387-393. Suave, N., Dzokoto A., & Opare B. (2002). The price of development: HIV infection in a semi urban community of Ghana. Journal of Acquired Immune Deficiency Syndrome (JAIDS), 20(4), 402-408. Swan, T. (1956). Economic growth and capital accumulation. Journal of Economic Record, 32(4), 334-362. Tahvonen, O., & Salo, S. (2001). Economic growth and transitions between renewable and non-renewable energy resources. European Economic Review, 45(8), 1379-1398. 124    Twerefo D. K., Akoena S. K. K., Egyir-Tettey, F. K., & Mawutor G. (2008). Energy consumption and economic growth: Evidence from Ghana. Paper presented at the Department of Economics, University of Ghana, Accra, Ghana. Weedy, B. M. (1987). Electric Power Systems (3rd ed.). Chichester: John Wiley & Sons Ltd. Wolde-Rufael, Y. (2004). Disaggregated industrial energy consumption and GDP: The Case of Shanghai, 1952-1999. Energy Economics, 26(1), 69-75. Wolde-Rufael, Y. (2005). Energy demand and economic growth: The African experience. Journal of Policy Modeling, 27(8), 891-903. Wolde-Rufael, Y. (2006). Electricity consumption and economic growth: A time series experience for 17 African countries. Energy Policy, 34(10), 1106- 1114. World Bank. (2011). World development indicators on online (WDI) database, Washington, DC: World Bank. Retrieved September 2011, from http://www.worldbank.org. Yang, H. Y. (2000). A note on the causal relationship between energy and GDP in Taiwan. Energy Economics, 22(3), 309–317. Yoo, S. (2006). The causal relationship between electricity consumption and economic growth in ASEAN countries. Energy Policy, 34(18), 3573-3582. Yoo, S., & Kwak, S. (2010). Electricity consumption and economic growth in seven South American countries. Energy Policy, 38(1), 181-188. 125    Yu, E. S. H., & Choi, J. Y. (1985). The causal relationship between energy and GNP: An international comparison. Journal of Energy and Development, 10(2), 249-272. Yu, E. S. H., & Hwang, B. K. (1984). The relationship between energy and GNP: Further results. Energy Economics, 6(3), 259-266. Yu, E. S. H., & Jin, J. C. (1992). Cointegration tests of energy consumption, income and employment. Resources and Energy, 14(3), 259-266. Zachariadis, T. (2007). Exploring the relationship between energy use and economic growth with bivariate models: New evidence from G-7 countries. Energy Economics, 29(6), 1233–1253. 126    APPENDICES APPENDIX A RESULTS OF PLOTS OF VARIABLE IN LEVELS 6.2 6.3 6.4 6.5 6.6 6.7 6.8 6.9 7.0 1975 1980 1985 1990 1995 2000 2005 2010 LY 20.8 21.2 21.6 22.0 22.4 22.8 1975 1980 1985 1990 1995 2000 2005 2010 LEC 8 12 16 20 24 28 32 1975 1980 1985 1990 1995 2000 2005 2010 FD 0 20 40 60 80 100 120 140 1975 1980 1985 1990 1995 2000 2005 2010 INF 51 52 53 54 55 56 57 58 1975 1980 1985 1990 1995 2000 2005 2010 LF Source: Generated by the author using Eviews 5.0 127    APPENDIX B RESULTS OF PLOTS OF VARIABLE IN FIRST DIFFERENCE -.8 -.6 -.4 -.2 .0 .2 .4 .6 .8 1975 1980 1985 1990 1995 2000 2005 2010 DLEC -.16 -.12 -.08 -.04 .00 .04 .08 1975 1980 1985 1990 1995 2000 2005 2010 DLY -10 -8 -6 -4 -2 0 2 4 1975 1980 1985 1990 1995 2000 2005 2010 DFD -120 -80 -40 0 40 80 120 1975 1980 1985 1990 1995 2000 2005 2010 DINF -.2 -.1 .0 .1 .2 .3 .4 1975 1980 1985 1990 1995 2000 2005 2010 DLF Source: Generated by the author using Eviews 5.0 128    APPENDIX C TEST FOR THE ORDER OF INTEGRATION (AUGMENTED DICKEY FULLER): LEVEL AND FIRST DIFFERENCE WITH INTERCEPT Levels (Intercept) 1st Difference (Intercept) Var. ADF-Statistic Lag Var. ADF-Statistic Lag OI LY -0.1367(0.9379) 1 DY -4.207(0.0021)*** 0 I(1) LEC -2.6463(0.1929) 0 DLEC -5.42159(0.000)*** 1 I(1) LK -1.17553(0.6754) 0 DLK -6.50816(0.000)*** 0 I(1) FD -1.1792(0.6737) 0 DFD -6.31189(0.000)*** 0 I(1) INF -5.29(0.000)*** 0 DINF -7.7464(0.000)*** 0 I(0) LF -2.1023(0.2450) 5 DLF -6.9535(0.000) *** 1 I(1) Note: D denotes first difference. *** and * represent significance at the 1% and 10% levels respectively. Numbers in brackets are P-Values. OI represents the order of integration. Source: computed using Eviews 5.0 Package 129    APPENDIX D TEST FOR THE ORDER OF INTEGRATION (AUGMENTED DICKEY FULLER): LEVEL AND FIRST DIFFERENCE WITH INTERCEPT AND TREND Levels (Trend & Intercept) 1st Difference (Trend & Intercept) Var. ADF-Statistic Lag Var. ADF-Statistic Lag OI LY -1.228207 (0.8901) 1 LY -5.275(0.0006) *** 0 I(1) LEC -2.67079( 0.1369) 1 DLEC -5.339(0.0005) *** 1 I(1) LK -2.579303(0.2913) 0 DLK -6.413(0.0000) *** 0 I(1) FD -1.631907(0.7617) 0 DFD -6.346(0.0000) *** 0 I(1) INF -5.6811(0.0013) *** 0 DINF -7.787(0.0000) *** 1 I(0) LF -2.117969 (0.5183) 4 DLF -5.042(0.0000) *** 3 I(1) Note: D denotes first difference. ***, ** and * represent significance at the 1%, 5% and 10% levels respectively. Numbers in brackets are P-Values. OI represents the order of integration. Source: computed using Eviews 5.0 Package 130    APPENDIX E TEST FOR THE ORDER OF INTEGRATION (PHILLIPS PERRON): LEVEL AND FIRST DIFFERENCE WITH INTERCEPT Levels (Intercept) 1st Difference (Intercept) Var. PP-Statistic Bwd Var. PP-Statistic Bwd OI LY -0.33065(0.9110) 3 DLY -4.16(0.0023)*** 3 I(1) LEC -2.167788(0.2208) 3 DLEC -4.74(0.0005)*** 3 I(1) LK -1.09707(0.7074) 2 DLK -6.61(0.0000)*** 4 I(1) FD -1.61679( 0.6388) 3 DFD -6.31(0.0000)*** 3 I(1) INF -5.40(0.0001)*** 4 DINF -14.66(0.000)*** 6 I(0) LF 1.92350(0.9997) 5 DLF -3.39(0.0074)*** 5 I(1) Note: D denotes first difference. ***, ** and * represent significance at the 1%, 5% and 10% levels respectively. Numbers in brackets are P-Values. OI represents the order of integration. Bwd represents bandwidth. Source: computed using Eviews 5.0 Package 131    APPENDIX F TEST FOR THE ORDER OF INTEGRATION (PHILLIPS PERRON): LEVEL AND FIRST DIFFERENCE WITH INTERCEPT AND TREND Levels (Trend & Intercept) 1st Difference (Trend & Intercept) Var. PP-Statistic Bwd Var. PP-Statistic Bwd OI LY -1.258618(0.8834) 3 DLY -5.21 (0.0007)*** 3 I(1) LEC -2.732248(0.2300) 3 DLEC -4.66 (0.0032)*** 3 I(1) LK -2.618067(0.2749) 1 DLK -6.50(0.0000) *** 4 I(1) FD -1.686961(0.7381) 3 DFD -6.36 (0.0000)*** 2 I(1) INF -5.68(0.0002) *** 1 DINF -15.6(0.0000)*** 8 I(0) LF -2.117969(0.5183) 4 DLF -5.04(0.0000) *** 3 I(1) Note: D denotes first difference. ***, ** and * represent significance at the 1%, 5% and 10% levels respectively. Numbers in brackets are P-Values. OI represents the order of integration. Bwd represents bandwidth. Source: computed using Eviews 5.0 Package 132    APPENDIX G PLOT OF CUMULATIVE SUM OF RECURSIVE RESIDUALS 2010 15 10 5 0 -15 20082003199819931988198319781973 -10 -5 The straight lines represent critical bounds at 5% significance level Source: Generated by the author using Microfit 4.1 133    APPENDIX H PLOT OF CUMULATIVE SUM OF SQUARES OF RECURSIVE RESIDUALS 1.5 1.0 0.5 -0.5 0.0 1973 1978 1983 1988 1993 1998 2003 2008 2010 The straight lines represent critical bounds at 5% significance level Source: Author’s computation using Microfit Version 4.1 134    APPENDIX I DATA USED FOR THE STUDY year LF INF FD M D LK LEC LY 1971 51.521 9.5588 18.032 1 0 2.5207 5.7371 22.726 1972 51.387 10.067 20.287 0 0 2.1595 5.8398 22.701 1973 51.293 17.683 20.867 0 0 2.0353 5.9629 22.729 1974 51.232 18.135 19.297 0 0 2.4774 5.9811 22.796 1975 51.196 29.825 22.631 0 0 2.4526 5.9240 22.663 1976 51.183 56.081 25.176 0 0 2.2859 5.9587 22.627 1977 51.197 116.45 22.151 0 0 2.2405 5.9853 22.650 1978 51.238 73.092 19.486 0 0 1.6215 5.7923 22.731 1979 51.302 54.441 19.621 0 0 1.9064 5.9800 22.705 1980 51.389 50.070 16.210 1 0 1.8079 6.0432 22.710 1981 51.497 116.50 13.756 1 0 1.5524 6.0250 22.674 1982 51.621 22.296 15.541 0 0 1.2617 5.9195 22.603 1983 51.759 122.88 9.6841 0 1 1.3247 5.0976 22.556 1984 51.906 39.665 9.7515 0 1 1.9247 4.5257 22.639 1985 52.063 10.305 11.468 0 1 2.2544 5.0865 22.687 1986 52.229 24.565 11.325 0 1 2.2295 5.5461 22.739 1987 52.405 39.815 11.735 0 1 2.3380 5.6535 22.786 1988 52.591 31.359 12.414 0 1 2.4194 5.7092 22.841 1989 52.785 25.224 13.927 0 1 2.5769 5.7217 22.891 1990 52.988 37.259 13.312 0 1 2.6663 5.7780 22.923 1991 53.196 18.031 13.376 0 1 2.7615 5.8166 22.975 1992 53.409 10.056 17.002 1 1 2.5444 5.8604 23.013 1993 53.633 24.960 17.347 1 1 3.1691 5.8174 23.060 1994 53.872 24.870 18.635 1 1 3.1168 5.8179 23.093 1995 54.130 59.462 18.380 1 1 3.0507 5.8576 23.133 1996 54.409 46.561 17.696 1 1 3.0106 5.8930 23.178 1997 54.702 27.885 20.194 1 1 3.1712 5.9609 23.219 135    136    Data used in the study continued 1998 55.002 14.624 21.156 1 1 3.1074 5.6185 23.265 1999 55.298 12.409 21.652 1 1 3.0189 5.8366 23.308 2000 55.581 25.193 23.214 1 1 3.1398 5.8003 23.344 2001 55.845 32.905 25.767 1 1 3.3004 5.8187 23.384 2002 56.104 14.816 29.305 1 1 2.9325 5.7408 23.428 2003 56.343 26.675 28.119 1 1 3.1328 5.4177 23.478 2004 56.566 12.625 29.216 1 1 3.3456 5.3946 23.533 2005 56.773 15.118 29.494 1 1 3.3674 5.5079 23.590 2006 56.965 10.915 19.981 1 1 3.0744 5.6979 23.652 2007 57.144 10.733 22.255 1 1 3.0011 5.5101 23.715 2008 57.310 16.522 23.596 1 1 3.0658 5.5864 23.796 2009 57.461 19.251 25.795 1 1 2.9741 5.5801 23.835 2010 57.601 10.708 26.2041 1 1 3.10928 10.1915 23.9091 Source: world Bank, (2011).