RESIDENTIAL ENERGY DEMAND MODELING IN DEVELOPING REGIONS - THE USE OF MULTIVARIATE STATISTICAL TECHNIQUES

This paper presents a new approach for modelling residential energy demand. In developing regions it is considered a new approach as energy demand equations apply to 'homogeneous' groups of consumers which are endogenously defined. The structural analysis of energy demand is the first phase of the approach which defines the set (tree) of consumption groups. These are obtained by applying multivariate statistical techniques on cross-section pooled data. Principal components analysis and discriminant analysis are the main methods used. The set of groups obtained is linked to a set of equations through a qualitative response model which simulates the decisions of households. Equations concerning energy consumption, the choice of energy equipment and the repartition by energy products are then estimated for each group. An application for the Cyclades region is presented. The model consists of 269 equations, 261 parameters and 15 exogenous variables. Scenarios and results for 1985-2000 are given together with a number of suggestions for further research.