EMF WP 12.3: A Structural Comparison of Models Used in EMF 12 to Analyze the Costs of Policies for Reducing Energy Sector CO2 Emissions

As part of every energy modeling forum (EMF) study, a comparison of model structure is conducted in order to better understand the (often disparate) results reported by the models. In EMF 12, this effort has been extensive due to the heterogeneity of the fourteen participating models.

This diversity has a number of advantages, chief among them being the greater number of insights that can be obtained by comparing the models’ results in various scenarios which are relevant to the global-warming policy debate. To fully appreciate both individual model outputs and intermodel output differences, however, it is necessary to understand how each model’s structure affects its results. The disadvantage of a diverse set of models is that it complicates the structural comparison, and hence the interpretation of results. This paper develops a paradigm which facilitates the process.

The fundamental policy issue with which EMF 12 is concerned is the economic cost incurred when policies are enacted that reduce energy sector carbon emissions. Since the energy sector is the dominant carbon emission source, and the demand for energy is closely tied to the demand for goods and services throughout the economy, the interpretive framework is based upon a simple identity relating domestic energy sector carbon emissions to gross domestic output (GDP) and domestic energy consumption. Each term in the identity is decomposed until the rate of exchange in carbon emissions is a function of price and quantity change in each sector of the economy, including the secondary and primary fuel markets. Although this model cannot itself be used to generate emissions reduction scenarios, it helps focus the structural comparison on those aspects of the participating models which affect the results of interest.

These model characteristics fall into three general areas of comparison. The first is the general model type, e.g. recursive general equilibrium or dynamic optimal growth, which also includes the models’ time horizons, data sources, and foresight assumptions. Here it is seen that, despite the many differences among the fourteen models, there are five broad classes within which models actually have striking similarity. Moreover the class in which a model fits indicates the types of policy questions the model is fit to answer.

Understanding the distinctions between these five classes also simplifies the analysis in the remaining two areas of comparison: demand modeling and supply modeling. These last two sections are forthcoming.