This research seeks to examine the relationship between energy consumption and economic growth in Tanzania. The study will investigate the causal relationship on how energy consumption affects economic growth in Tanzania. With the existing literature indicating that there is a positive causal relationship between economic growth and energy consumption, aggregate and disaggregate data on electricity, energy and oil will be used to examine the relationship between energy consumption in Tanzania between 1990 and 2017. To achieve this, Granger causality and ARDL boundary approaches will be employed. In the first part, a general understanding of global energy consumption will be will be discussed briefly. Second, data on annual energy consumption and economic growth series of Tanzania between 1990 and 2017 will be examined. Our contribution will be to provide policymakers with a new dimension of approaching economic growth through increasing energy consumption.
Keywords: Tanzania, energy consumption, economic growth, panel cointegration, panel causality.
Global energy consumption overview; global primary energy consumption increased by just 1% in 2016, following growth of 0.9% in 2015 and 1% in 2014, this compares with the 10-year average of 1.8% a year (BP; June 2017). Global energy consumption has about doubled in the last three decades of the past century. In 2004, about 77.8% of the primary energy consumption is from fossil fuels (32.8% oil, 21.1% natural gas, 24.1% coal), 5.4% from nuclear fuels, 16.5% from renewable resources, of which the main one is hydroelectric, 5.5%, whereas the remaining 11% consists of noncommercial biomasses, such as wood, hay, and other types of fodder, that in rural-economies still constitute the main resource Beretta, G. P. (2007).
Africa energy consumption overview; Africa energy consumption; in 1991, per capita consumption of modem energy in Africa was estimated to be 12 GJ which is less than half that of South-America and less than one-tenth that of Europe (World Resources Institute, 1994c). Despite the fact that African Continent possesses the potential environment for energy sources for energy production but the energy consumption in general and electricity consumption, in particular, is very low (Karekezi and Kimani, 2002; ECA, 2004). Therefore there is less average in use for energy consumption compared to an average person used energy in England more than a century ago (Davidson and Sokona, 2002). Even more glaring is the wide disparity within African countries themselves. For instance, in Ghana, 62% of the urban population has access to electricity while only 4% of the rural population has access to electricity (Saghir, 2002). Electrification rates range from as low as 3.7% in Uganda, 4.7% in Ethiopia and 5.0% in Malawi to as high as 45% in Ghana, 50% in the Ivory Coast and 66% in South Africa (IEA), 2002). Similarly, electric power consumption per capita ranged from as high as 556kWh in Zambia, 698kWh in Gabon and 845kWh in Zimbabwe to low as 22kWh in Ethiopia, 47kWh in the Democratic Republic of the Congo and 58kWh in Tanzania (World Bank, 2003). The average per capita electricity consumption for Sub-Saharan Africa (excluding South Africa) was 112.8kWh in 2000, representing a mere 5% of the world average.2 With only 23% of its population electrified compared to the world average of 73%, Africa has the lowest electrification rate of any major world region (IEA, 2002). Although the African continent has 14.1 percent of the world’s total population lives in but, the continent consumes only 4.2 percent of world processed energy for industrial uses in 2007 (IEA, 2010).
Tanzania Energy consumption Overview; Tanzania has diverse energy sources that are untapped, the energy sources include biomass, hydro, uranium, natural gas, coal, geothermal, solar and wind 10. Odhiambo, N.M. (2009). The primary energy includes biomass (90%); petroleum (8%); electricity (1.5%), and the remaining (0.5%) is contributed by coal and renewable energy sources. About 80% of the energy that delivered from biomass is consumed in rural, while the importation of oil costs about 25% to 35% of the nation’s foreign currency earnings Msyani, C. M. (2013). To date, only about 18.4% of the country’s population has gained access to electricity. Extending the National Grid to many parts of the country including rural areas is not financially and economically feasible, Msyani, C. M. (2013). The high rate of dependence on biomass energy indicates that the government has not paid as much attention to energy development as it should be. This may be one of the reasons for the very slow rate development process in the economic region. It causes some concern not only for the government perspective but also for environmentalists owing to continuing use of biomass has adverse consequences on soil fertility and the environment in general Goreau, T. J., & de Mello W. Z. (1988) .The Annual Energy Consumption for Tanzania is 5,740.84GWh (2012) The Highest Energy Demand Stands at 16.9 GWh/Day. Only 14% of the Country is Electrified (12% of Urban And 2% of Rural) Access to electricity is about 18.4%. Current Total Number of Customers Is 1,032,000 Maximum Number of Connections per annum achieved is 90,000.
Problem Statement and justifications; Pattern of energy consumption in Tanzania is being affected by the energy prices. The serious problems of fuel and increasing energy consumption have brought the focus of Tanzania on the role of energy in economic performance. However, the growing concerns over energy scarcity and environmental costs of energy have attracted the interest of the government in Tanzania to declared that a variety of energy strategies are implemented to promote the rational and efficient use of energy. Thus, the causal relationship between energy and economy has undergone investigation. Whether energy consumption leads to economic growth or economic growth stimulates energy consumption have been examined in a number of studies. The causality in either direction between energy consumption and economic growth may have a significant impact on energy-saving policies. If causality runs from economic growth to energy consumption in a country, this indicates economic of the country is less dependent on energy. Energy saving policies may have no adverse effect on economic growth. However, if causality runs in the reverse direction, this suggests an energy-dependent economy in which a shortage of energy may adversely affect income (Narayan and Smyth, 2008). The economic growth and energy consumption are highly dependent and energy conservation measures may negatively affect economic growth (Asghar, 2008). Hence, the study to investigate the causality relationship between economic growth and energy consumption is vital so that the energy conservation policies may be pursued without adversely affecting the Tanzanian economy.
The significance of the study findings: The findings from the study will help to contribute to reform Tanzanian energy strategy efforts towards advanced energy consumption and consequently curb environmental problems and foster economic growth. Therefore Policy makers will make use of the findings from this study to devise short-term, medium-term and long-term strategies for sustainable natural resource management.
Why focus on the energy consumption and economic growth? : At the advanced level, both energy consumption and economic growth are on consumers as the ultimate target. While economics deals with the allocation of scarce resources among consumers’ competing want Wood TS, Baldwin S (1985), people’s welfare is the central concern of the economic growth h systems.
Why study the household sector? : The household sector consumes the greatest proportion of total energy across the country. In Tanzania, the household sector accounts for 80-91% of total energy consumption in the country Hysen B (2011). Statistics further reveal that in sub-Saharan Africa (SSA) household cooking alone takes up to 60-80 percent of the total national energy use . In SADC region, households consume 97 per-cent of wood energy for cooking, heating and cottage industries 14. Andrea B, Goldemberg J (1996).
Different studies have pointed out factors that affecting energy consumption that related to economic growth especially they focus on fuel accessibility, fuel affordability, fuel reliability, fuel flexibility, Oil prices, household type and effective household size, climatic conditions, dwelling technology technics and ownership, stock of liquid assets (wealth); future income expectation, urban-rural location differences, and level of consumer indebtedness.
Objectives of the study
The specific objectives of the study were to:
a. Analyze patterns of energy consumption and economic growth in Tanzaniab. Analyze factors affecting energy consumption and economic growth in Tanzania c. Investigation of households’ preferences to wood fuel from natural forests.Hypotheses of the study
This study puts forward the following main hypotheses; the feedback hypothesis suggests that energy consumption and growth are interrelated and complement each other.
Literature survey on energy consumption and Economic growth
The hypotheses mentioned above have motivated scholars to put more effort for investigating the causal relationship between energy consumption and economic growth. Therefore, some studies attempt to examine the causal relationship between energy consumption and economic growth. The previous researches definitely based on time-series data of specific countries and, apply the Engle and Granger residual-based cointegration test (1987) and its maximum test based on Johansen (1988) and Johansen and Juselius (1992). For example, by employing Granger causality test, Ebohon (1996) shows the causal relationship between energy consumption and economic growth. The previous advanced time series in last decades, analysis techniques have evolved and the energy consumption and economic growth relationships are carried out by using the Toda and Yamamoto tests of Granger causality (1995). For example, Wolde-Rufael (2005) investigate the long-run relationship between per capita energy use and per capita GDP for 19 African countries and finds mixed results, ranging from negative causality to bidirectional causality. Akinlo (2008), by using the Autoregressive Distributed Lag (ARDL) bounds and Granger causality tests based on Vector Error Correction Model (VECM), explores the causal relationship between energy consumption and economic growth for 11 Sub-Saharan African countries and finds mixed results. He reveals that economic growth and energy consumption are cointegrated and, there is a bidirectional relationship between energy consumption and economic growth for 3 countries and a unidirectional causality running from economic growth to energy consumption for 2 countries. The “neutrality hypothesis” for the energy–income relationship is confirmed in respect of 5 countries. With the same method, Odhiambo (2009b) finds a unidirectional causality running from energy consumption to economic growth in Tanzania. Wolde-Rufael (2009), in a multivariate framework including labor and capital as additional variables and, by using Granger causality test of Toda and Yamamoto, re-examines the causal relationship between.
Granger causality (Granger, 1969) can be used to analyse the extent that change of past values of one variable account for the later variation of other variables. Usually, Granger causality exists between variables and if by using the past values of variable the variable can be predicted with a better accuracy, and relating to a case when past values of variables are not being used, with an assumption that other variables stay unchanged. Granger causality test usually analyses two variables together, testing their interaction. Gelo, T. (2009).. All of the possible permutations of the two variables are:
• Unidirectional Granger causality from variables to variables,
• Unidirectional Granger causality from variables, to variables
• Bi-directional casualty,
• No causality.
In all conditions, the possible common assumption is that the data are stationary. Stationary of a data in random Process indicates that its statistical property does not change (constant) with time. If the Granger causality not in non-stationary time data can lead to false casual relation (Cheng, 1996). Economic and energetic time series usually have some challenges of non-stationary series. This is due to the fact that most often lies in the constant change of legal and technical regulations and rules, and is making changes in the economic relations, which causes the change of time series. Infarct the change of regulations can affect the stationary time series, but in that case the relation between variables before and after the changes is stable. Non-stationary time series are trying stationary with certain mathematic procedures, for example, differentiation of variables.
Granger test of causality analyzes if the equation
= + + ?
Where 0 ? i, j ? T
Gives better results than equation:
= 0 (the null hypothesis,).
If the hypothesis is rejected where = = … = = 0, than it can be implied that according to Granger causality causes variable. The statement which implies that x according to Granger does not cause y, is gained if the current value if x better explains the current value y, and the past values of x and y, than just past values of y. Granger causality test explains which variable is dependent and which is independent in the equation, and in the energy economics most often the long term relation is formed between energy consumption and income of a country, and it is expressed through the gross domestic product. According to Granger (1986), the test is valid if the variables are not cointegrated
The second important element is the analysis of lag length. The result of Granger causality test is very sensitive to the selection of lag length. If the chosen lag length is less than the true lag length, the omission of relevant lags can cause bias. If the chosen lag length is more the true lag length, the irrelevant lags in the equation cause the estimates to be inefficient and does not give expected results Gelo, T. (2009). If they share common trends i.e. they have long-run equilibrium relationships thus two or more variables are said to be cointegrated. But this technique of cointegration involves three steps such as Determination of the order of integration of the variables of interest. For this purpose two popular tests are used: namely Dickey Fuller (DF) and Augmented Dickey Fuller (ADF) test that is based on expanded Dickey Fuller test; Perron (Perron 1988), Philips and Perron (1988) made a PP test, which is considered more useful for aggregates data; Kwiatkowski et al. (Kwiatkowski at al., 1992) KPSS test; Perron (Perron 1989) with PB test which is considered more useful when there structural breaks time series data with PB test, which is considered better than other tests, if there are structural breaks in time series data Gelo, T. (2009).
Combining these tests, there are four different results to be considered:
1. Rejection with ADF and PP tests and the acceptance with KPPS test offer firm proof stationary of the analyzed data
2. The acceptance of ADF and PP tests and the rejection of KPPS test offers firm indication I(1)
3. Acceptance of all the tests indicates that the data with insufficiently long series of data is not representative enough.
4. Rejection of all the tests indicates that series of data is not I (1) nor I (0).
In normally the Dickey-Fuller (DF) test and Augmented Dickey-Fuller (ADF) test (Dickey and Fuller 1979, 1981) are commonly applied on: is not I (0), which are given by the following equations:
(DF) ? = a + b +
Where denotes the variables GDP, energy consumption (total energy consumption or energy consumption of specific c for, of energy like electricity, oil, gas…). All the variables are real and in log form. ? is the difference operator, a and b are parameters to be estimated.
(ADF) ? = a + b + ? + ADF test:
a, b, and care the parameters to be estimated, y is element t. The tests are based on the null hypothesis (): is not I(0), If the calculated DF and ADF statistics are less than their critical values from Fuller’s table, then the null hypothesis () is rejected and the series are stationary or integrated. In the second step co-integration between variables is estimated using variables:
– Engle and Granger technique (Engle and Granger 1987), or
– Johansen maximum likelihood approach (Johansen 1988; Johansen and Juselius 1990, 1992)
The co-integration equation estimated by the OLS method is given as:
= + +
Where and are the income and energy consumption.
In the third step residuals () from the cointegration regression are subjected to the Stationarity test based on the following equations:
(DF) ? = ? + +
(ADF) ?= ? + t ? +
Where, is the residual from equation gained by OLS method. If b is negative and the calculated DF or ADF statistics is less than the critical value from Fuller’s table, the null hypothesis of non-stationarity is rejected. On the other hand, if the null hypothesis of non-stationarity is rejected and the variables are not cointegrated then the standard Granger causality test is appropriate.
In the third step vector, error-correction modeling and exogenous variables test are used. Engle and Granger introduced a new method for the analysis of time series in 1987. The assumption for their modeling is that they are stationary. Time series is a stationary, if its arithmetic means does not depend on time, and if its variance does not change systematically through time. That implicates that the value of variance is a definite number. Therefore, time series return to the middle of the series and fluctuate around it, around its constant range. In practice, that mostly is not the case. Time series can be transformed, but working with such series leads to cases where it is difficult or almost impossible to interpret gained results. By overcoming such circumstances, Engle and Granger have proven that if the independent series is integrated by the sequence I(d), and if the residual linear regression are among these variables integrated by the same order, I(d-b), then the series are coin grated sequences d,b, CI(d,b). In order to detect integration, it is necessary to note the order of the integration of variables x and y. Non-stationary series is causing problems when unit rooted, which equals them, being integrated into the first order. Such series are series of a random walk, according to which the future value equals the past value increased by error. Random walk series are difficult for predicting future. Therefore, it is necessary for them to be tested for unit roots, and it is necessary to discover the order of integration. Causality in econometric relates to the possibility of one variable, predicting (and therefore causing) the rest of the variables. The relation between these variables can be described by the VAR models. In this case is possible that variable influences, that influences, as well as there exists mutual influence of these variables, or that these variables are nondependent of each other. Granger causality test comes down to the estimation of following VAR mode
With the assumption of being correlated and producing white noise. All the variables, used in analysis have unit roots, approximately of 5% of significance. Non-stationary has been removed by differentiation.
The hypothesis for Granger casualty test are:
…. Does not influence
Data used Total primary energy consumption (TPEC) and the fluctuation of economic activity (GDP) are connected the observed variables. Therefore total primary energy consumption is the result of the consumption graph of particular forms of primary energy (coal and coke, liquid Fuels, natural gas, hydropower, biomass, electricity, waste and renewable).The variables will be used to display the facts of the cumulative graphs that depict the report of energy consumption and economic growth in Tanzanian country.
Results The finders will be obtained due to the realizing connection between the energy and GDP, the first Granger test has been conducted, that applies to the relation of the total primary energy consumption, and gross domestic product. Also, the Review can be used to analyze the estimated data.The econometric test is used to determine whether economic growth affecting the energy consumption or energy consumption affect the economic growth.
The model of analysis will be estimated and later the Granger causality test will be taken as shown from the following models.
VAR Model is:
Log= + logpe + log
Log= + loge + log
The hypothesis for the Granger causality test are:
…. Does not influence,
The Variable GDP is logarithmic, and the original GDP series had unit roots lggdp has differentiated and also being transformed into lggdpd if. The total primary energy consumption variable is logarithmic, lgtpec, and since it had unit roots it as well differentiated, lgtpecsd if was gained therefore a test of the unit root without the differentiation for the total primary energy consumption. Also, the P-values can be used to determine the conclusion values which leads to acceptance or rejections. The significance or insignificance values of gained or rejections of the results can be seen in the fact that GDP is the cause of the change of total primary energy consumption or increasing the primary energy consumption that is normally being caused due to the change in GDP.
From the hypothesis concept detriment, the relation between energy consumption how it affects the economic growth in Tanzanian by increasing or decreasing or it tends to exist proportional. The result that obtained using the Granger causality tests and cointegration analysis normally indicate some relationship for Tanzania from primary energy consumptions to gross domestic products, or from gross domestic products to primary energy consumption.