Analysis of the Determinants of Electricity Demand in Libya - Research Proposal Example

Empirical study on the determinants of demand for electri in Libya, 1980 (A dissertation proposal) I. Introduction A. Background Energy isone of the main and most important elements for achieving social and economic development. It is one of the most essential sources of energy. The importance stems from the role of electric power in fulfilling main consumption needs of basic energy requirements as one of the inputs to the production process. This study aims to analyze and understand the determinants or factors affecting the demand for electricity in Libya during the period (1980-2010). The study will focus on the use of ordinary least square techniques in constructing the empirical model of the demand function for electricity. B. Rationale and relevance of the study One dominating theme in economic theory is that the economy is always in equilibrium or moves towards an equilibrium. Although there are several economic perspectives that attempt to contradict the dominant theme, the theory of an economy at equilibrium or moving towards equilibrium continues to be the main conventional perspective in economics. Assuming, that equilibrium indeed prevails or that the economy always moves towards equilibrium, we can argue therefore that the demand function can suffice in estimating the consumption of electricity. In any case, even if the economy fails to move or be in equilibrium, the demand function remains useful in determining society’s needs for electricity. While as per one construction of the Say’s Law in that supply does not create its own demand, assessing demand remains important as societies and economies exist precisely to satisfy human want. Thus, estimation of demand is extremely important. Further, estimating the parameters of the demand function for electricity has a practical relevance. It will allow policy makers or business to anticipate the need for power plants in Libya. The Libyan government will find the concern relevant because it will allow them to anticipate needs even before awareness of the needs emerge. The business sector is also concerned with Libya’s electricity needs because it is both a source of opportunities and it plays a role in enhancing the viability of the Libyan business sector (Libya Energy Report, 2003). In at least one time, the US Commercial Service has recognized the business potential of Libya (US Commercial Service 2006, p. 43). From the standpoint of theory development, determining a valid or non-spurious demand function for electricity will allow us to define the key variables influencing demand for electricity in Libya. C. Problem Statement What is the demand function for electricity in Libya? How and in what magnitude do price, income, population, appliance imports, and temperature affect the demand for electricity? How accurate does a demand function based on the said variables track the realized demand for electricity in Libya? How can we improve the accuracy of the said demand function in tracking realized demand? How do we compare the tracking ability of the said demand function for electricity in Libya vis-à-vis other models that purport to track the consumption of electricity in Libya? Does an empirical or econometric model of demand for electricity perform a better job in tracking electricity consumption in Libya? D. Objectives 1. Develop an econometric or empirical model of demand function for electricity in Libya and improve the said model based on empirical findings and analysis 2. Compare the tracking ability of the demand for electricity model with the other models that can be developed for Libya 3. Assess the implications of findings for policy and direction for study E. Scope The scope of the study will be 1980 to 2010 during which data are available. Although the study focuses on the construction of an econometric model of electricity consumption based on the demand function, the study will also assess the tracking ability of the model and will compare the said ability vis-à-vis alternative models. This work will also assess the implications of its findings on policy. The policy implications, for instance, could cover implications related to production, planning relative to Libyan oil reserves, exports, and perhaps even on the prospects for utilizing renewable energy in Libya. II. Review of Literature We begin with the more recent works. Abosedra et al. (2009) estimated the demand for electricity using ordinary least square (OLS), autoregressive integrated moving average method (ARIMA) also known as the Box-Jenkins methodology, and exponential smoothing for January 1995 to December 2005. In the Abosedra et al. study, the ARIMA model, followed by exponential smoothing and OLS, was most accurate in tracking demand for electricity during the period. Based on their study, the authors recommended univariate time-series modelling for forecasting demand “until detailed data on various socio-economic variables becomes available” (Abosedra et al. 2009, p. 11). The said Abosedra et al (2009) findings indicate that the authors are also dissatisfied with their ARIMA model. ARIMA modelling uses the lag, averaged, and differenced values of electricity demand for forecasting future demand (Gujarati 1995, p. 735-746; Abosedra et al 2009, p. 13) and, thus, ARIMA models do not provide a good theoretical foundation for forecasting other than a claim that demand may be following a business cycle or seasonal fluctuations. Abosedra et al.’s exponential smoothing (2009) does not provide information as well as the demand for electricity was merely regressed using a time trend variable and variables to represent seasonal fluctuations. The exponential smoothing technique indicated only two things: 1) demand follows a linear time trend; and 2) demand fluctuates across seasons. Unfortunately, the Abosedra et al. (2009) study gave a poor rating for the OLS technique. However, the Abosedra et al. (2009) OLS estimate on the demand function for Lebanon can be described as good given that their OLS regression for electricity demand for Lebanon has an adjusted R-square of 0.823038 implying that variations in the values of the regressors explain around 82% of the variations in the regressand or dependent variable which is demand for electricity in Lebanon in this case. The Abosedra et al. (2009) regression of Lebanon’s demand for electricity used real imports, relative humidity, and monthly temperature as regressors. Despite a better theoretical foundation for the regression, however, Abosedra et al. (2009) gave their OLS regression a poor rating. It is the position of this work that Abosedra et al.’s (2009) OLS have a better foundation in theory and that Abosedra et al. (2009) should have perfected the model instead of rating the OLS as the inferior regression relative to the ARIMA and exponential smoothing methodology. Neeland (2009) studied demand for electricity in the United States for 1970 to 2007 using the Augmented Dickey Fuller (ADF) unit root test, Johansent test, and a rolling regression. Although the Neeland study did not provide fit statistics of the regression, the author claimed nevertheless that his regression showed that the “the primary driver of adjustments in electricity consumption is the own price elasticity of demand and growth in real income per capita” (2009, p. 193, 200). In the words of Neeland (2009, p. 202), the study “determined that changes in electricity price and real income per capita, more than any other facto examined, influence energy consumption patterns of US residents”. The other variables examined by Neeland included temperature change and the price of a substitutes. The ADF unit root test applied by Neeland used the change in prices as the dependent variable and the lagged and differenced values of price (Neeland 2009, p. 196). The Johansen test involved testing for the co-movement of variables (Neeland 2009, p. 196). Finally, the rolling regression involved a “simple linear regression” in which the “earliest observation is dropped and the latest is added” (Neeland 2009, p. 196). Unfortunately, data on fit and R-squared are not available on the Neeland (2009) article that can allow us to make a good assessment of the Neeland regressions. The work of Labandeira et al. (2009) estimated the demand for electricity in Spain using data from September 2005 to August 2007 using observations from 422,696 households, 30,499 companies and 688 large consumers (p. 9). Labandeira et al. (2009) used a demand function for companies and households. For the households, the independent variables covered climatic factors, and household characteristics. Dummies for time and location were also included to explore possible location and temporal effects on electricity consumption. There were missing data and Labandeira et al. (2009, p. 9) handled the situation by executing transformation techniques. For household demand for electricity, a log linear model of the variables was employed thereby making possible the computation of elasticity values. Labandeira et al. (2009, p. 15) implicitly expressed that in the future, a possible improvement in the model can factor in the role of decision to consume goods linked to the use of energy product. This matter is one of the concerns of this study. Unfortunately, the Labandeira et al. (2009) regressions did not report statistics that would enable us to assess the validity of their regressions. Report on the adjusted R square and goodness of fit data are not available on their regressions. Further, statistics on possible serial correlation were not reported by Labandeira et al. (2009). Nevertheless, Labandeira et al. (2009, p. 11) concluded from their regressions that electricity demand is inelastic to price in the period assessed. Labandeira (2009, p. 11) also reported that income and activity are important factors in explaining the demand for electricity among households. In a concern that this proposed study seeks to address, the Labanderia et al. (2009) regressions also found that temperature changes have a small but highly significant effects on consumer demand. For Brazil, Carlos & Notini (2009) applied a time varying parameter error correction model to estimate the country’s demand for electricity during the period 1997 to 2007. However, despite their sophisticated modelling, the adjusted R-squares of their two regressions are poor values 0.352707 and 0.059238. This is very different from the value obtained by Abosedra et al.’s OLS adjustred R-square of 0.823038. On this basis, the Carlos & Nossini (2009) regressions are inferior compared to the Abosedra et al. (2009) regression that included an OLS regression. For Mexico, Chang & Chombo (2009) used co integration and error-correction models with time varying parameters for estimating the demand for electricity. Despite their sophisticated models, their best regression has a value of 0.829. Although one of the Chang & Chombo (2009) regressions has an adjusted R squared value of 0.975, the regression has Durbin Watson statistics of 1.47 indicating serial correlation and low regression credibility. Earlier for Spain, Villadongos (2006) used an ordered probit with instrumental variables of contracted power by consumers and found that electricity power contracted by each household depends on electricity consumption and household characteristics. The household characteristics include ownership of central heating, location of community, rural/urban characteristics, year of construction of the dwelling unit, size of dwelling unit, gas supply, and type of house ownership (Villadongos 2006, p. 14-15). The maximum likelihood R squared of the Villadongos (2006) regression, however, is low at 0.334 (p. 16). Meanwhile, a study done a year earlier than Villadongos (2006) used OLS techniques for estimating demand for electricity based on the 1973 to 1995 data and found that home ownership is a relevant factor that explains spending for energy in Spanish households (Labandeira et al. 2005, p. 17). Based on the regressions, Labandeira et al. (2005, p. 17) also found that demand for electricity is negatively related to the educational level of the head of the household, positively related to income, qualitatively related to age, and qualitatively related to place of residence. Labendeira et al. (2005) articulated that their work is a significant contribution to literature in that their regression demonstrated the role of household characteristics in the demand for electricity. Unfortunately, however, while t-ratios for the Labandeira et al. (2005) regressions are available, the F, adjusted R-squared, and Durbin Watson statistics are not. Guertin et al. (2003, p. 33-34) highlighted the role of appliances, temperature, and house characteristics in influencing demand for electricity in Canada. For the United Kingdom, Patrick & Wolack (2001) developed framework for estimating demand for electricity in the United Kingdom and found that demand for electricity can vary within a day. Patrick & Wolack argued that their findings could be used by firms to develop price bid policies that are variable even within a day (p. 42-45). In sum, Patrick & Wolack argued that their paper have formulated a general framework for estimating the within-day customer demand for electricity. III. Methodology A. Theory and Model Development Frank (2003, p. 44-45) affirms that the determinants of demand are income, tastes, prices of substitutes and complements, changes in expectations, and population. Frank (2003, p. 47) pointed out that economists describe a “change in demand” to mean a shift in the entire demand curve. On the other hand, a “change in quantity demanded” refer to a movement along the demand curve when price changes (Frank 2003, p. 457). The discussion of Frank (2003) is of long standing and well-established in economics. For instance, Frank’s 2003 discussion echoes the discussion on demand and the distinction between changes in demand and movement along the demand curve in several editions of the book of Richard H. Leftwich on “The Price System and Resource Allocation” such as the old Leftwich (1979). Leftwich (1979, p. 32) the demand function as: (1) x = f (px, T, C, I, pw, R, E) Where, according to Leftwich (1979, p. 32): x = quantity of good x px = price of good x T = taste or preferences C = population of consumers I = consumer’s income and distribution of income pw = price of related goods R = range of goods and services available E = market expectations. In this work, however, we construct the demand function for electricity as: (2) Qt = f (Pt, Yt, Nt, IMt, Dt, Qt-1) Equation (1) can be transformed into a log linear model by taking natural logarithms (Ln) of both sides (Greene 2008, p. 13): (3) LnQt = a0 +a1LnPt +a2LnYt +a3LnNt +a4LnIMt +a5LnDt +a6LnQt-1 +ut. Where: Qt = electricity demand at time t and expressed in thousands of megawatt-hours (MWH). a0 = a constant term. a = coefficients of independent variables, where i=1,2,......,6. Pt = average electricity price at time t in Libyan dinars. Yt = real income (expressed in real Gross Domestic Product GDP). Nt = number of population at time t. IMt = value of imported appliances at time t. Dt = difference between the average maximum and the average minimum yearly temperatures at time t. Qt-1 = Qt lagged one year. ut = the error term, with the usual assumptions. Greene (2008, p. 13) described the preceding log linear equation as the constant elasticity equation because the equation implies that the partial derivative of Q with respect to each regressor is constant. There is good reason to use the log linear form. Greene 2008 (p. 13) pointed out that the log linear is often used in models of demand and production function. The approach that this study hopes to use was informed by several studies discussed in the review of literature. Based on the review of literature, this work has come to believe that the OLS is just as good as the other regression techniques for estimating parameters and can have a good tracking ability or accuracy that is comparable with the other regression estimation techniques. At the same time, interpretation of OLS results can be relatively easy and forthright especially with regard to variable relationships. From the regression of Abrosedra et al. (2009), this proposed study was informed that imports and temperature change can have an effect on the demand for electricity. The proposed study learned from Neeland (2009) that real income, temperature, and the price of substitute goods influence demand for electricity. Labandeira et al. (2009) was confirmatory on the role of temperature change in affecting demand for electricity while Guertin (2003) highlights the role of appliances and temperature on affecting demand for electricity. B. Main Hypotheses The applicable hypotheses are as follows: 1. Ho: price of electricity have no affect on demand for electricity H1: price of electricity negatively affect demand 2. Ho: real income have no effect on demand for electricity H1: real income positively affect demand for electricity 3. Ho: population have no effect on H1: population positively affect demand for electricity 4. Ho: imported appliances have no effect on demand for electricity H1: imported appliances positively affect demand for electricity 5. Ho: temperature change have no effect on demand for electricity H1: temperature change positively affect demand for electricity C. Variables to use 1. For estimation purposes, the 1980 to 2010 data set available in Libya and several authoritative international agencies shall be used. 2. Both the regressand and regressors will be in natural logarithm format consistent with the log linear form earlier discussed. 3. The role of price of electricity in affecting demand will be investigated primarily through its Libyan dinar value. 4. The role of role income on demand for electricity will be assessed primarily through the real GDP that may be inflated or deflated through an appropriate reference year to ensure comparability across all years. 5. The role of imported appliance will be investigated through the value of imports inflated or deflated based on a base year consistent with that used for computing the real GDP. 6. Temperature shall be measured based measured either in degrees Celsius or degrees farenheit. D. Analysis Data analysis will follow the usual techniques for regression analysis. The log linear format allows analysis related to elasticity. Regression diagnostics will be used to assess the validity of regression results. Tests of cointegration and Granger causality will be used to check against spurious regressions. Further, the regressions obtained by this study will be compared with other demand function regression models to assess its reliability. References Abosedra, S., Dah, A., & Gosh, S. (2009). Demand for electricity in Lebanon. International Business & Economics Research Journal, 8 (1), 11-18. Carlos, A. & Notini, H. (2009). Brazilian electricity demand estimate: What has changed after the rationing in 2001 (An application of time varying parameter error correction model). Paper presented before the 21st Meeting of the Brazilian Econometric Society. Available from http://virtualbib.fgv.br/ocs/index.php/sbe/EBE09/paper/view/993/345 [Accessed 17 March 2010]. Chang, Y. & Chombo, E. (2003). Electricity demand analysis using cointegration and error-correction models with time varying parameters: The Mexican Case. Working Paper 2003-08. Houston: Department of Economics, Rice University. Greene, W. (2008). Econometric analysis (6th Ed., International). New York: Pearson Prentice Hall. Gujarati, D. (1995). Basic econometrics (3rd Ed.). New York: McGraw-Hill, Inc. Guertin, C., Kumbhakar, S., & Duraiappah, A. (2003). Determining demand for energy services: Investigating income driven behaviours. Manitoba: International Institute for Sustainable Development. Frank, R. (2003). Microeconomics and behaviour (5th Ed.). Boston: McGraw Hill. Labandeira, X., Labeaga, J., & Rodriguez, M. (2005). A residential energy demand system for Spain. Madrid: Center for Energy and Environmental Policy Research. Labandeira, X., Labeaga, J., and Otero, X. Estimation of elasticity price of electricity with incomplete information. Documento de Trabajo 2009-18. Madrid: Fundacion de Estudios de Economia Aplicada. Leftwhich, R. (1979). The price system and resource allocation (7th Ed.). Illinois: The Dryden Press. Libya Energy Report (2003). Executive summary: Oil, refining and petrochemical, gas, electricity, water. ________: Norton Publications. Neeland, H. (2009). The residential demand for electricity in the United States. Economic Analysis & Policy, 39 (2): 193-203. Patrick, R. & Wolak, F. (2001). Estimating the customer-level demand for electricity under real-time market prices. Working Paper 8213. Massachussets: National Bureau of Economic Research. US Commercial Service (2006). Doing business in Libya: A country commercial guide for US companies. Villadangos, L. (2006). Pricing household electricity demand in Spain: Equity and efficiency. Paper presented, Simposio de Analisis Economico, Spain. Available from http://www.webmeets.com/SAE/2006/Prog/viewpaper.asp?pid=244 [Accessed 17 March 2010]. Read More