V. Times Series, and Econometric and Judgmental Forecasts
1. Time-series models
2. Judgmental forecasts and econometric models
3. Combined forecasts
VI. Summary and Conclusions
Appendix I. Rules for Pricing of Crude Oil Futures Contracts
Appendix II. The Relationship Between the “West Texas” and Other Crude Oil Prices
1. Test of Unconditional Unbiasedness: Full Sample
2. Test of Unconditional Unbiasedness: Increasing and Decreasing Spot Prices
3. Comparison of Futures Prices and Random Walk
4. Comparison of Alternative Weighting Schemes for Futures Prices
5. Comparison of “MINDIS” Weighted Forecasts
6. Comparison of ARMA Model and Futures Prices Accuracy
1. Crude Oil: Spot Prices and Excess Returns
2. Crude Oil: Spot Prices and One-Month-Ahead Forecasts
3. Crude Oil: Spot Prices and Three-Months-Ahead Forecasts
4. Crude Oil: Spot Prices and Six-Months-Ahead Forecasts
5. Crude Oil: Spot Prices and Nine-Months-Ahead Forecasts
6. Weighting Schemes
1. Crude Oil: Size of Market (New York Mercantile Exchange)
2. Comparison of Alternative Weighting Schemes for Futures Prices: Absolute Mean Forecast Errors
3. Crude Oil Prices and Differentials, 1988-90
4. Crude Oil Prices and Differentials: Correlation Matrix
5. Instrumental Variable Estimates
This paper analyses the development of futures markets in crude oil and examines the accuracy of forecasts obtained using futures prices. Futures markets in crude oil have grown extremely fast during the last five years, and the volume of trade in futures transactions far exceeds the trade in the spot market. The depth and breadth of the futures markets suggest that forecasts obtained from futures prices are unlikely to be biased and are likely to provide a relatively accurate indication as to the future course of spot prices.
A number of empirical exercises are undertaken to evaluate the “unbiasedness” hypothesis, and the accuracy of the forecasts. An extensive dataset, covering the period from the inception of crude oil trading on the New York Mercantile Exchange to 1990, is utilized for this purpose. An analysis of the mean excess returns that could be obtained from holding futures contracts did not appear to suggest any systematic bias in the futures prices. This result complemented the results of the comparison of forecasts using futures prices with forecasts using a random walk model, which showed that the former provided more accurate forecasts for all forecast horizons. As the length of the forecasting horizon increased, however, the accuracy of both types of forecasts diminished markedly.
An analysis of intra-month futures prices suggested some marginal improvement in forecasting accuracy for distant horizons, compared with the end-of-the month prices. When weights on intra-month prices were determined endogenously, it appeared that the weighting scheme should be related to the length of the forecast horizon. Futures-prices forecasts were also more accurate compared with forecasts obtained from time-series models as well as judgmental and econometric forecasts. Combining forecasts from alternative techniques, however, yielded only a marginal improvement in terms of variance of forecast errors.
The empirical results strongly suggest that futures prices provide forecasts that are, in general, superior to those obtained from alternative techniques for short term horizons. For more distant horizons, their accuracy does diminish markedly; however, even for these horizons the futures forecasts are no worse, and are often better, compared with those obtained from alternative techniques.