words, few existing models can incorporate subnational policies or analyze the incidence of different cap-and-trade program designs, but both are important for understanding federal options.

Identifying and Quantifying Uncertainty

Sensitivity analyses are useful to identify key areas of uncertainty within the modeling community. They are especially important for the largest models, to examine critical issues in some isolation, such as the Env-Linkages model did recently to look at carbon leakages. Janet Peace of the Pew Center on Global Climate Change did note that sensitivities tend to be conservative and do not often consider efficiency improvements or technology development occurring more rapidly than one would ordinarily expect. National models like NEMS require more sensitivity to international experiments. A more comprehensive analysis, the Monte Carlo simulation, is difficult for the large and complex models, even with low-cost computer time. John Reilly noted that MIT recently completed a Monte Carlo analysis of the EPPA model (Webster et al., forthcoming). IGEM is experimenting with the Delta method to exploit the standard errors in parameters and the standard error in, for example, the estimated energy input demand function, to try and create confidence intervals around the outcomes it produces—an approach that might be applicable to other large-scale models.

Representing Behavior

Building on discussions from the previous panel, several workshop participants remarked on the importance of behavior and how it is represented in the various models. Marilyn Brown pointed out that as social preferences evolve, there may be generational differences that are not currently captured by modeling, such as a preference for walkable (i.e., not car-dependent) communities, or a preference for locally produced foods. Tom Kram noted that the Netherlands Environmental Assessment Agency has embarked on a study to examine the relevance of dietary preferences, specifically moving from an animal-based diet to a more vegetarian-based diet. Many participants remarked that absence of a capability for modeling evolving preferences is currently a major limitation of most available analytical tools. Richard Goettle wondered whether the functional forms in the models actually represent the right way to look at the world. Models focus on things like maximizing utility, but that may not reflect real-world household decisionmaking.

John Conti noted that NEMS tries to incorporate behavior in its estimates, by going beyond least-cost modules. Dallas Burtraw stated that the demand side of the electricity equation is where the Haiku model faces its biggest challenges now. One area is the role for efficiency, and attempts have been made to address this by looking for opportunities on a broad national scale, and combining those with incentive-based programs to promote energy efficiency. He emphasized that consumers consume electricity services, not kilowatt-hours, so how does incentive-driven behavior affect outcomes? Resources for the Future is now trying to model the effects of time-of-day pricing in the electricity sector.

Peter Evans pointed out that international relations research has suggested that as countries interact, they may not act “rationally” in the economic sense by pursuing absolute gains, and instead might be concerned with relative gains. Several other participants noted that analyses are only beginning to take this in to account. On a similar point, Tom Kram explained that regional differences among countries in terms of preferences can be significant, but these are not captured in global models.

Systems Interactions

One important insight that IAMs provide in particular is human-Earth systems interactions. While this linkage makes detailed regional or inter-temporal results difficult, it does offer several advantages,



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement