Uses and Abuses of Marginal Abatement Supply Curves
The objective of the workshop’s first session was to discuss the proper interpretation and use of marginal abatement supply curves, which chart the cost of reducing greenhouse gas emissions through the deployment of various technology and policy measures. For each measure under consideration, its marginal cost is plotted against the net associated emissions reduction, and the results are stack-ranked from lowest to highest cost to form the marginal abatement supply curve. Marginal cost supply curves have been in use for decades, and a 2007 report released by McKinsey & Company represents a recent application to the study of reducing greenhouse gas emissions (McKinsey & Company, 2007). Marginal abatement supply curves are often used to link the results of bottom-up engineering analyses of the cost and technical potential of technologies with top-down economic models that assess the macroeconomic and energy system impacts of reducing greenhouse gas emissions. However, embedded within such supply curves are critical assumptions, including the baseline against which the supply curve is built (which may not be internally consistent across the specific technology options included in the supply curve), cost assumptions concerning the technologies represented within the supply curve, discount rates, and even assumptions concerning how rapidly or easily technologies might be deployed. Yet these assumptions may not be apparent to analysts who incorporate such supply curves into their models, or to policy makers who use a model’s results in making policy decisions. Further, a McKinsey-type supply curve that represents a broad array of technology options gives the illusion that all options have an equal probability of implementation, face no deployment constraints, and benefit from specific policies and measures identified to spur deployment, and that all lower-marginal-cost options would be exhausted before a move to the next least costly option. Such were the issues that provided motivation for this workshop session.
Issues in the use of energy conservation and greenhouse gas abatement cost curves were first discussed by Mark Jaccard of Simon Fraser University, who began his talk with a description of energy efficiency cost curves and greenhouse gas abatement cost curves. He described the possibilities offered by technology options with lower life-cycle costs (i.e., offering cost savings) that have been shown to have negative costs, meaning that the more efficient replacement technology has a life-cycle cost lower than that of the technology it replaces. Figure 2.1 shows an example of a cost curve associated with different options for reducing electricity consumption. Mapping electricity rates, one could make an argument that any of the efficiency measures, those steps on the curve in Figure 2.1 that are below electricity rates, would represent profitable actions for people to take on a private cost basis. Figure 2.1 also shows that, if society is looking at making an investment in a new supply option like a new hydropower dam, the cost of that option can be mapped on the curve and the result used to show that efficiency
actions below the cost of a new hydropower dam would be socially profitable compared to building the dam. Jaccard described it as basically the same methodological thinking that leads to carrying the supply curve approach from a focus only on energy efficiency to a focus on greenhouse gas abatement. Efficiency cost curves were popular 30 years ago, and greenhouse gas abatement cost curves have been around for at least 20 years. But Jaccard noted that leading energy-economy modelers have moved away from the supply curve approach, arguing that the curves mislead about costs and are unhelpful with policy. Jaccard believes that is probably too strong a statement and, as someone who comes from both an economics and a technology engineering background, he expressed his belief that there is useful information in such curves and in developing hybrid approaches, while still remaining cognizant of the issues with these curves.
Jaccard focused on several issues he sees as problematic with such supply curves. The first is that the construction of cost curves implies that each action is completely independent of every other action, for example, that installing efficient light bulbs is independent of making building shells more efficient. It also assumes that market conditions are homogeneous such that the cost of deploying the first 20 percent of the technology is the same as the cost of deploying the last 20 percent. Finally, the curves assume that a new technology is a perfect substitute and that the quality of service and the risks of adopting a new technology are identical to those associated with the technology being replaced. Responses to these issues have involved modelers constructing integrated models that have energy supply and demand working simultaneously and tracking within the models different vintages of equipment stocks. Such models can also portray the heterogeneous character of market responses and estimate the behavioral parameters that explicitly or implicitly incorporate nonfinancial values such as preferences related to technology attributes. He noted that models that are technologically richer or more explicit about technologies are more often called hybrid models, and these models have algorithms that simulate how people, firms, and households choose technologies. Jaccard argued that, although these models and their parameters are highly uncertain, research on technology deployment tends to focus on them because of the general awareness of the limitations of simple supply curve approaches.
The final point in Jaccard’s talk concerned the relevance of traditional supply curves for policy and what can be done to improve their use. He stated that the implicit message from traditional cost curves is that it seems
very inexpensive to achieve substantial reductions in energy use or greenhouse gas emissions. Such a message can suggest to policy makers that, if the costs are so low, there is no need for the kind of compulsory policies that really change market incentives, such as emissions pricing and regulations. He recommended instead the use of integrated hybrid models to construct marginal abatement cost curves in which each point on a curve has simultaneous actions occurring in an equilibrium solution (for example, adoption of more efficient lighting occurs with improvements in building shells, and their interactions are represented), a particular action (such as use of more efficient light bulbs) occurs continuously along the curve, and that the curves incorporate intangible costs and estimated responses to policy.
The second speaker in the session was Jayant Sathaye, the head of the International Energy Studies Program at the Lawrence Berkeley National Laboratory, who discussed empirical insights possible for energy-climate modeling from efficiency (supply) cost curves. Sathaye reminded the workshop audience that efficiency cost curves were developed about 30 years ago to enable a comparison of the potential and cost of energy efficiency options with supply-side potential and costs. He discussed several issues associated with the individual energy-reducing technologies and measures represented within the cost curves: (1) the baseline against which individual savings are measured; (2) the barriers to deploying these technologies or implementing these measures; (3) the program costs needed to implement and possibly subsidize the adoption of an energy-saving measure; and (4) the time frame during which a measure is effective. Sathaye noted that capturing all the issues that impede the full deployment of the energy-reducing measures in the cost curve would produce a curve showing about 45 percent of the savings that would be estimated without including these impacts.
Sathaye went on to discuss the impacts of incorporating non-energy benefits into curves and how such benefits become very important for the industrial sector. Besides reductions in energy costs, there may be reductions in atmospheric emissions of non-greenhouse-gas pollutants, generation of liquid and solid waste materials, and operations and maintenance costs. Sathaye pointed out that reductions in energy use alone will not cause most industries to purchase efficient technologies. Including non-energy benefits can greatly alter the cost curves, in some instances significantly increasing a technology’s cost-effectiveness. Figure 2.2 indicates the potential impact
of including potential non-energy benefits in the supply curve for the U.S. steel industry. Including non-energy benefits also can greatly alter the ranking of the technologies in terms of their relative benefits.
The final issue brought up by Sathaye was that efficiency cost curves are constructed as though they are static in time. However, it is known that over time costs drop for various energy-saving technologies in the industrial sector, as well as in the residential and commercial sectors. Sathaye cited steel making, residential gas furnaces, and commercial air conditioning equipment as specific instances in which costs have fallen as energy efficiency has risen. Thus, cost curves should evolve over time, and this issue should be considered when applying these curves.
The remainder of the session included a panel discussion and comments from the audience. The four discussants were Marilyn Brown, a professor at the Georgia Institute of Technology and a member of the workshop planning committee; Rich Richels, head of the Climate Division at the Electric Power Research Institute and also a member of the workshop planning committee; Howard Gruenspecht, deputy administrator for the EIA; and Hillard Huntington, a professor at Stanford University and the executive director of Stanford’s Energy Modeling Forum. Marilyn Brown talked about some of the ways that supply curves can be advanced to better reflect the ability of policies to make a difference in the marketplace. To address some of the concerns raised earlier in the session about the limitations of technology supply curves, Brown recommended the construction of policy supply curves that represent bundles of technologies that would be deployed in response to a policy. Figure 2.3 shows an example of such a curve from a recently released project (Brown et al., 2010). Policy supply curves allow multiple technologies to be modeled—for example, in the case of residential building codes a number of different advances and technologies that can be utilized to meet a code. Brown also noted that such curves are amenable to the inclusion of program administration costs.
Richels began by noting that the efficiency supply curves produced by the McKinsey study, echoing many studies from the early years (the late 1980s) of the climate change debate, showed many no-cost and negative-cost
options that nevertheless omitted additional hidden costs. The current goals for mitigation are such, Richels felt, that the policy debate should not be about whether there is a free lunch in mitigating climate change, but rather about whether the lunch is worth paying for. He expressed the concern that the debate over “how many free $20 bills are lying on the sidewalk” is irrelevant and should not be used as an excuse for policy paralysis. Hillard Huntington recalled that most of the same issues discussed earlier in this workshop session had been brought up more than a decade ago in an Energy Modeling Forum activity on supply curves. Despite some interesting things that have to be done analytically, Huntington was convinced that it is very important to communicate with policy makers about how to use these curves and the factors that change the shape and cost-effectiveness of these curves. He noted that behavioral issues appear to be critically important to explaining the gap between the technology opportunities and other energy-saving measures shown within these curves and the adoption of these measures by individuals and companies. Howard Gruenspecht began his remarks by concluding that the presenters and commenters had made it clear that analysts need to sharpen their focus on behavior in a variety of dimensions when assessing the costs of reducing energy use and greenhouse gas emissions. He went on to note that his agency’s (EIA’s) models include some behavior and a lot of technology detail. The EIA models use a mixed approach whereby decisions in some sectors are benchmarked to past behavior, whereas in other sectors, such as electric power generation, decisions are assumed to be based on a pure cost-minimizing behavior. He noted that recent experience suggests too little emphasis might have been placed on behavioral considerations, even in the electric power sector.
The session ended with comments and questions from the audience. Richard Moss from the Pacific Northwest National Laboratory/University of Maryland’s Pacific Joint Global Change Research Institute wondered whether the debate has moved beyond whether there are negative cost opportunities ($20 bills on the sidewalk) to the question of how we can use policy to more economically and efficiently bring about some of the transitions necessary to address climate change. Further, Moss noted that many of the claims made about different policies leading to job creation or improvements in energy security do have an economic component to them and yet are really difficult to get our hands around. He wondered how it might be possible to build on such studies of bottom-up technical potential for reducing energy use and emissions, and move onto some of these other challenging questions. Marilyn Brown responded by noting a growing appreciation that the market is not operating effectively, that intervention can improve things, and that many of the policies in place actually present barriers to efficient decision making. These barriers include the coupling of profits by the electric utility industry and the gas industry to the amount of revenue obtained, which discourages policies that reduce electricity or energy consumption. Rich Richels responded by recommending greater transparency in packaging some of the work that is being done, citing a talk he had heard recently about green jobs that mentioned only the number of jobs that would be added by adopting certain renewables, and did not discuss the potential negative impacts on other segments of the employment market. Richels’ conclusion was that, unless you give the whole picture, you are setting yourself up for being discredited.
Ed Ryder from Dow Chemical brought up the point that, although supply curves provide an entry point for discussion, one of the issues from an industrial perspective is the competition for capital and whether you spend your limited resources on energy efficiency projects or on some other projects that allow you to meet other objectives such as producing products in greater volume, expanding into different regions of the country or the world, or spending in another manner that provides greater returns on investment. William Nordhaus from Yale University noted that many of the comments on supply curves have been scornful of the bottom-up engineering approaches that are used to estimate the technical potentials shown in these curves. What he finds very exciting for the next decade or two of research is to bring to bear some of the important new advances in behavioral economics or the behavioral sciences more generally on issues related to supply curves.