A prerequisite for developing the planning and demandmanagement policies for energy is the appropriate projections of futuredemand. Capacity targets may be set particularly more efficiently in thelight of a knowledge of the probability of shortages. However, suchprobabilities are usually not available. Such probability distributionsare generated in this study. More specifically, a simple four-equationmodel has been developed to make projections of average and peak demandfor the city of Karachi in the year 2000. Three of the four equationshave been estimated using Full Generalised Least Squares withPrais-Winsten transformation in order to correct serially correlatederrors. The estimated model, after making appropriate tests forhetroscedasticity, has been put to recursive bootstrapping to generateprobability distributions of average and peak demand in order to assessthe extent of uncertainty in point projections. Bootstrapping has beenused because of the limitations of the conventional method with regardto imposing a pre-specified stochastic structure on the error term andassuming the knowledge of the values of the exogenous variables for theforecast period with certainty. Probability distributions are based on1000 random samples. The results indicate that under quite plausibleassumptions, the extent of uncertainty remains significant which shouldbe taken into account for future policy planning.