dc.description.abstract | Maria Wist Langmoen
Own-price and income elasticities for household electricity demand -A Literature survey using meta-regression analysis
Economists have been modelling the electricity demand for several decades in order to explore the effect of governments taxation policies, to predict future demand and reveal needs for investment, and see the effect of price fluctuations on the demand. There is a wide range of studies to be found in the area, both because electricity demand can easily be measured, the interesting particularities of this demand and the practical importance of the results. The problem is that the results are not homogeneous. In the articles used for this review the income elasticities could be found in an interval of [0,04; 2,09], and own price elasticities [-0,91; 0,51]. When it comes to estimating effects of a new tax on the demand of electricity, the wide range of elasticity estimates will cause problems for policy makers that want to establish the effects the tax will have on the electricity market.
The Meta-Analysis, "the analysis of analysis" cf. Glass (1976), proposes a method to compare results from different, but comparable, studies to create a consensus on different empirical subjects in social sciences. This master thesis is formed as a literature review of empirical articles on the demand for electricity, using meta-analysis as a fundament to comment the heterogeneity of the elasticity estimates. I estimate the expectation of the own-price and the income elasticity subject to a set of variables describing the original article's estimation procedure and the environmental context of the estimation.
The work in this master thesis precedes a review of elasticities from all over the world (project in Statistics Norway), which will include a wider number of articles on residential electricity demand. Meta-analyses will also possibly be performed on elasticities in other fields. The main objective of this master thesis is to go through the methodology and see which problems one encounters both with building a databank and with the theory related to meta-analysis. This will be done through the performance of a meta-analysis with the 16 studies related to the electricity demand.
The Meta-regression analyses have been used for other research topics than economics. In medicine and psychology for example, several studies on the same topic have been gathered to compare the results, synthesize and find general tendencies. I follow the background theory developed by Stanley (2001) and Florax et al. (2002), which both contribute to the methodology on how to carry out a meta-analysis in economics. Florax (2001) proposes to use the meta-analyses as a tool for calibrating elasticities and other measures for macro-economic modelling. In particular elasticities and correlations are economic measures that are possible to model in a Meta-analysis because they do not depend on the unity of measure. First I use Stanley (1989) and Florax et al. (2002) approach in a Ordinary Least Square estimation. And further, following the tools developed by statisticians for studies in other social sciences (used in pedagogy, sociology and medicine), I correct for heteroscedasticity by using the standard deviation recuperated from the studies (Hunter and Schmidt (1990) and Hedges and Olkin (1985)). These standard deviations will be used in general least square estimations on the same parameters.
I start by describing the Meta-analysis methodology for economics, and how it can be used to integrate findings across studies. The methodology and assumptions are related to the Ordinary Least squares, the Weighted Least squares and the Generalized Least squares regressions.
Then the reader is introduced to the economic demand theory related to household electricity demand. Three main points of view are presented in a brief manner in order to explain how a difference in theoretic background could affect the elasticity estimates. Further in the fourth section I describe the concept of elasticity, in order to familiarize the reader with this measure.
Section V describes the choice of variables that could be causing the difference between the estimates of own-price and income elasticities related to residential electricity demand. These were included in the meta-analysis. The section VI give a short introduction on problems related to my dataset, and problems related to creating a more extensive data set.
Section VII contains parameter estimates of the different regressions along with comments relate to these variables and analysis of residuals. As the GLS estimation results did not seem to be specified appropriately according to the variance in the original studies I chose to comment only the OLS and WLS results. From these regressions I conclude that the Own-price elasticity depends in a statistically significant way of the demand model specification, the data set and inclusion of the trend variable. The Total Expenditure model gives a significantly higher own-price elasticity estimate and macro data gives a lower estimate. The length of the time series in the data contributes to a higher elasticity. When it comes to the estimation of the income elasticity the results are not as interesting, and I conclude that this could be due to misspecification. The income elasticity estimate depends significantly of, if the temperature and the stock of appliances is included in the original model.
As I agree with Hunter and Schmidt's statement that I quote in page 6, section II, I think the integration of findings across studies will be of growing importance in the future. Not only because of the increasing amount of studies that are done on different areas in economics, but because of the globalisation. The opening of markets and the need for creation of new regulations make it important to have good knowledge about the different economic relationships. | nor |