This paper presents results of a modeling study undertaken to quantify the effects of recent changes in energy prices on future total energy consumption and oil's market share. To predict these two quantities vs time, five pairs of regional models, which together span the Free World, were developed. The explanatory variables in the models are income, population, total energy price, crude oil price, and non-oil energy price. The most recently published data were used to estimate coefficients in these models. Projections of total energy demand and oil share through 2000 for different scenarios are given. The sensitivity of future demand for crude oil to changes in income and prices is examined. Both multiple regression and linear programming were used to estimate coefficients. These two methods gave essentially identical results under identical conditions. Because the LP approach allows greater flexibility in imposing judgemental bounds on coefficients and selectively weighting historical data, it was used to obtain the coefficients presented.
We find that considerable judgement must be applied to obtain coefficients that cause the models to yield reasonable results. The tailored computer system developed to use the models allows hindcasting as well as forecasting. Hindcasting experiments reveal that a perfect history match does not always mean that the model will provide reasonable projections.