A compact model of oil supply disruptions

Abstract

We have developed a model of oil price jumps caused by oil supply disruptions. The core of the model is a compact general equilibrium model of oil demand. Given an exogenous forecast of oil supply and potential GNP for the world, the model can forecast the oil price, real GNP, and value added for the world. The data base for the model consists of historical time series of world oil supply, world oil price, and growth rates for world GNP. The data demonstrate that small changes in oil supply are associated with large changes in oil price. If a large change in oil price causes a small change in oil consumption, the short-run price elasticity of demand must be small. We have specified a model with a short-run and a long-run price elasticity. The six parameters in the model have been estimated using historical data for three cases. The initial model had five parameters. For Case 1, we used a search procedure to determine the parameters that minimized the root mean square (RMS) of the differences between the price backcast by the model and the historical data on oil price. We found that the RMS error was 121% and that the price calculated by the model was too low for the period from 1974 to 1980. After a review of the historical data on oil consumption, oil price, and world GNP, we concluded that the response of the world oil market to the 1974 oil price shock was different than the response to the 1979-80 oil price shock. After the jump in oil price from 1973 to 1974, the consumption of oil decreased in 1975 but quickly recovered and reached a peak in 1979. After the jump in oil price from 1978 to 1980, the consumption of oil declined steadily for four consecutive years. To improve the model's capacity to simulate the historical data, we introduced a technological change factor that increases the demand for oil in the period from 1971 to 1979. For Case 2, the rate of technological change is 3.8% and the RMS error is 35%. For Case 3, the rate of technological change is 7.0% and the RMS error is 29%. Although Case 3 has the smallest error, we concluded that the Case 2 set of parameters provided the best match for the historical data on world oil price. For Case 2, we find a small short-run elasticity ( − 0.09), a substantial long-run elasticity ( − 0.92), and a lag of 0.08, which corresponds to 12 years. We have invented an oil price scenario that is similar to the historical data and have determined the corresponding oil supply scenario for the Case 2 set of parameters. We have investigated the impact of an increase or decrease of five million barrels per day in 1998.