Abstract
M.Com. (Econometrics)
An attempt has been made in this study to model electricity sales of Eskom with the aid of an econometric model. The study examines the history of econometric modeling and the applications thereof. These applications include forecasting, policy simulation and policy analysis. The sales of electricity within the specific sales categories is estimated and simulated with the use of an econometric model. The model is specified
according to the expected dependent and explanatory variables. The a priori
theoretical considerations concerning the size and sign of the parameters of the
function are also included in the specification, as well as the mathematical form
of the model. The specification process of the econometric model is based on
econometric theory and on available information relating to the phenomenon
being studied. The method of ordinary least squares is used in the estimation of the parameters of the model. As this is an econometric study, the emphasis is on the
evaluation of the results. Economic a priori criteria, statistical criteria and
econometric criteria are used to evaluate the results of the parameters obtained
by the method of ordinary least squares. Several tests, including the Goldfeld Quandt
test for heteroscedasticity, the Durbin-Watson test for auto correlation and the Frisch analysis of multicollinearity are executed. The overall results of the tests to which the model was subjected, was satisfactory. The best functions are combined in a structured model. This model is simulated with the use of the Gauss-Seidel-method and forecasts of historical values are obtained. Statistical tests for the validation of these results, as well as Theil's inequality coefficient are applied to test the forecasting power of the model. The results of the ex post forecast for the period 1988.1 to 1995.1
emphasize the usefulness of the model as a forecasting device and the dynamic
simulation demonstrates the ability of the model to reproduce the historical data
from which it is estimated.