A comparative analysis of errors in long-term econometric forecasts
The growing body of literature that documents forecast accuracy falls generally into two parts. The first is prescriptive and is carried out by modelers who use simulation analysis as a tool for model improvement. These studies are ex post, that is, they make use of known values for exogenous variables and generate an error measure wholly attributable to the model. The second type of analysis is descriptive and seeks to measure errors, identify patterns among errors and variables and compare forecasts from different sources. Most descriptive studies use an ex ante approach, that is, they evaluate model outputs based on estimated (or forecasted) exogenous variables. In this case, it is the forecasting process, rather than the model, that is under scrutiny. This paper uses an ex ante approach to measure errors in forecast series prepared by Data Resources Incorporated (DRI), Wharton Econometric Forecasting Associates (Wharton), and Chase Econometrics (Chase) and to determine if systematic patterns of errors can be discerned between services, types of variables (by degree of aggregation), length of forecast and time at which the forecast is made. Errors are measured as the percent difference between actual and forecasted values for the historical period of 1971 to 1983.
- Research Organization:
- Pacific Northwest Lab., Richland, WA (USA)
- DOE Contract Number:
- AC06-76RL01830
- OSTI ID:
- 6912671
- Report Number(s):
- PNL-SA-13921; ON: DE88013121
- Resource Relation:
- Other Information: Portions of this document are illegible in microfiche products
- Country of Publication:
- United States
- Language:
- English
Similar Records
Directory of Energy Information Administration model abstracts 1988
Directory of Energy Information Administration model abstracts