of drilling activity, quantities of oil and gas discovered, reserves of oil and gas, and production rates for a selected set of years. Whether or not individual models can calculate these intermediate variables is often important in communicating their capabilities, limitations, and dependability. Comparisons of model behavior at several observation points provide the raw material for group discussions and report writing.

Each model is run by its key developer or a close colleague. Third-party model operation is not routine. There are advantages in third-party model operation, but these are outweighed by the practical advantages of the modelers running their own systems. The developer may best understand the limits of applicability of the model. Moreover, the modeler knows which sets of equations or data within the model must be modified to examine specific scenarios. Having worked extensively with a given system, (s)he can make runs without undertaking the enormous learning cost required for third-party analysis. In addition, decentralized model operation supports the goal of enhanced communication among the model users and model developers.

The

Displays of model outputs are designed so as to facilitate interpretation and comparison. Graphic comparisons are prepared by the EMF staff for use by the working group. displays of the outputs, if done creatively, can help in interpreting the comparative behavior of the individual models. If done poorly, the display of results can hide more information than it communicates. Thus, this task is far more critical than may be apparent.

A large portion of working group time is devoted to the critique and interpretation of the runs. Differences among models provide an opportunity and a motivation for explaining why the different results occur. Discrepancies in results may point to fundamental differences in model structure, model parameters, basic data utilized, or perceptions about the direct implications of scenario assumptions. Divergence in model results normally leads to creative tensions among the modelers with each trying to understand why his model differs from the others. One motivation is to improve the model, if appropriate. Another is to show why one model's answers may be more dependable than those of another model. Divergences in results is a strong motivating force which leads to important understandings about the fundamental model differences relevant for policy or planning purposes, the areas of uncertainty in knowledge about the world, and the significant areas of research potential. The process has resulted in the revision of a model during the study to account for implementation problems not initially perceived by the model's developer.

These working group discussions enhance the model users' insights about the policy issues and suggest distinctions among policy options that may not be apparent on the surface. For example, some policy options may increase the cost of producing a given amount of energy, e.g.,

restrictions on types of coal that can be used for electricity generation or restrictions against the use of nuclear energy. Others, however, may simply influence the price of energy without influencing its cost, e.g., an energy tax that is redistributed through the U.S. Treasury. The first class will have a far greater impact on economic growth than will the second class of options [10]. What may seem to be a subtle distinction in policy options may profoundly impact effects of the various options on economic growth.

The working group report is characteristically written and published in two separate components: a relatively short summary, approximately 30 pages long, directed at a broad audience and a much longer series of supporting documents or chapters aimed at a smaller, more technical audience.

The summary explains the major commonalities and differences in the models, provides answers (to the extent possible) to the issues raised, identifies limitations of the analysis, and presents recommendations developed by the group. In writing this report communication is emphasized; the report is intended to be jargon-free and accessible to nontechnical readers. A two- to three-page executive summary encapsulates the key conclusions of the report.

The series of supporting documents varies significantly from study to study. Generally included, however, are descriptions of the individual models, comparisons of the models, a simplified framework for both comparing the models and communicating an intuitive understanding of the results, a detailed description of the scenarios, detailed results from each model with comparative graphics, and a set of technical papers discussing more deeply any modeling and analysis issues that may have surfaced during the study.

The working group report provides one communication vehicle for disseminating the results of individual studies. Other mechanisms also are used. The Senior Advisory Panel is briefed on the report and members have played significant roles in communicating the results. Working group observers bring insights back to their respective organizations and help to disseminate the study's findings more broadly. Working group members typically make seminar or conference presentations based upon the study. EPRI, as well as the EMF, publishes the report and facilitates its distribution. Thus, many individuals help publicize the study. The study belongs fundamentally to the working group and the communication of results relies heavily on study group members actively publicizing the results.

Although the EMF reports, individual participant presentations, and other vehicles are used to communicate the results of the study, much of its benefit is not easily transferred to nonparticipants. A major EMF focus is on how people can use models more effectively, but effective use is a skill, learned like any other skill. While an EMF study can help modelers and model users in the difficult, artistic process of utilizing models for addressing real, complex issues, the skills and insights gained often cannot be fully

EMF STUDIES

The products of the Forum* - comparative studies of significant energy issues include two completed studies, "Energy and the Economy" and "Coal in Transition: 1980-2000," and one virtually completed study, "Electric Load Forecasting: Probing the Issues with Models." "Aggregate Energy Demand Elasticities" is well under way; and a fifth study, "U.S. Oil and Gas Supply," is just beginning.

Table 2 summarizes the progress of the various studies as of March 1, 1979.

Two other study topics identified by the Senior Advisory Panel as high priority areas, "Energy and the Environment" and "World Oil Supply, Demand and Prices," are listed in Table 2. Although at this time it is uncertain whether either will be chosen for future study, the EMF staff is currently in the process of preliminary exploration and issue definition.

This section describes the studies, focusing primary attention on the study process and on the contribution of each to the evolution of the forum.

EMF 1: ENERGY AND THE ECONOMY

The Forum project was initiated in 1976 with a study designed to demonstrate the research concept. The working group used six models to study the nature and strength of the feedback from the energy sector to the aggregate economy, isolating the key factors determining the effect of energy system changes on the long-run economic growth. The results demonstrate the importance of the value share of energy in the economy, the flexibility in substituting other inputs for energy use, and the link between productivity and capital formation in explaining the behavior of the models [4].

A group of 30 model users and developers conducted the study. Because of the experimental nature of the project, William Hogan served as working group chairman as well as EMF Executive Director.

The six models each explicitly represented the energy-economic linkage. Each model was for the full U.S. economy, and each was judged appropriate for long-run issues but not for short-run issues. Common scenarios were constructed by standardizing many input variables. The working group then sought to explain the common results or the causes of model differences. This comparison process was facilitated by the high degree of commonality among the various models.

The key comparative results of the study were estimates of the aggregate elasticity of substitution** implicit in the participating models. This parameter was shown in the