Modeling Ontario regional electricity system demand using a mixed fixed and random coefficients approach

In examining the municipal peak and kilowatt-hour demand for electricity in Ontario, we begin by exploring the issue of homogeneity across geographic regions. A common model across municipalities and geographic regions cannot be supported by the data. We consider various procedures which deal with this heterogeneity and yet reduce the multicollinearity problems associated with regional specific demand formulations. The recommended model controls for regional differences assuming that the coefficients of regional-seasonal specific factors are fixed and different while the coefficients of economic and weather variables are random draws from a common population for any one municipality by combining the information on all municipalities through a Bayes procedure.