7. Comparing Cost Estimates for the Kyoto Protocol Introduction 86 This chapter provides a comparison of recent publicly available estimates of the costs of achieving the Kyoto Protocol carbon reduction targets in the United States for the period 2008 to 2020. The projections are compared for the years 2010 and 2020, when the information is available, for the following projection sources: the Energy Information Administration (EIA) using the National Energy Modeling System (NEMS). WEFA," Charles River Associates (CRA) using the MultiRegional Trade model (MRT), the Pacific Northwest National Laboratory (PNNL) using the Second Generation Model (SGM).87 the Massachusetts Institute of Technology (MIT) using the Emissions Prediction and Policy Analysis Model (EPPA), Electric Power Research Institute (EPRI) using the MERGE model and Data Resources, Inc. (DRI)." Differences between studies are related, to the extent possible, to the features of the modeling systems used (e.g., level of aggregation, level of geographic coverage), important assumptions employed, and the particular points of view embodied in the models."1 90 88 Two cases were solicited for analyses from each group: a 7-percent-below-1990 (1990-7%) case in which the United States is assumed to reduce carbon emissions to 1990-7% levels for the period 2008-2020 without the benefit of sinks, offsets, international carbon permit trading, or the Clean Development Mechanism (CDM): and a best estimate of the impact on U.S. energy markets if sinks, offsets, and Annex I emissions trading were allowed, but not global trading or CDM. Differences in the cost estimates for meeting the Kyoto Protocol targets can be related to important differences in assumptions about (1) economic growth in the reference cases without the Kyoto Protocol, (2) the status of the resources available (e.g., resource base, world oil prices, and the slate of technologies available to the marketplace). (3) the sensitivity of energy demand to price changes, (4) the degree of foresight that decisionmakers have in the marketplace. (5) the structure and function of the economy (e.g., how quickly the economy can shift to less energy-intensive industries when the price of energy relative to capital and materials increases), (6) the degree and speed of substitution for factors of production (capital, labor, energy, and materials) when their relative prices change, and (7) the representation of technology (ie., representation of vintaged energy equip. ment and the penetration of new technologies). Summary of Comparisons Because the information available varies considerably, a detailed comparison among the sources is virtually impossible. Therefore, a comparison of common variables is provided in this section, with an explanation for the differences between the sources. Comparisons are provided for three of the cases analyzed in this report: the 1990-7% case and two cases-9 percent above 1990 (1990+9%) and 14 percent above 1990 (1990+14%)-that are comparable in some respects to the Annex I trading case. The variables compared are carbon price, change in actual gross domestic product (GDP) from the respective reference case in each study. tute. 85 WEFA, Inc., Global Warming: The High Cost of the Kyoto Protocol, National and State Impacts (Eddystone, PA, 1998). 86 Both the CRA and WEFA studies have been supported to some extent by Industry groups, including the American Petroleum Insti 87 J.A. Edmonds et al., Modeling Future Greenhouse Gas Emissions: The Second Generation Model Description (Washington, DC: Pacific Northwest National Laboratory, September 1992). Runs using PNNL's SGM model formed the basis for the testimony provided by Dr. Janet Yellen, chairman of the Council of Economic Advisers, on March 4, 1998, before the House Commerce Committee, Energy and Power Sub committee. 88 H.D. Jacoby. R. Eckhaus, A.D. Ellerman, et al. "CO, Emission Limits: Economic Adjustments and the Distribution of Burdens," Energy Journal, Vol. 18, No. 3 (1997), pp. 31-58. MIT's analysis is part of a much larger integrated assessment methodology funded by the Office of Energy Research, U.S. Department of Energy. 89 A.S. Manne and R.G. Richels, "On Stabilizing CO, Concentrations-Cost Effective Emissions Reduction Strategies," Energy and Environmental Assessment. Vol. 2 (1997), pp. 251-265. EPRI's work is self-funded and is part of the research agenda of electric utilities. 90 Standard and Poors DRI, The Impact of Meeting the Kyoto Protocol on Energy Markets and the Economy (July 1998). 91 Information used in this chapter was contributed by Dr. Montgomery and Dr. Bernstein of Charles River Associates, Dr. Richels of the Electric Power Research Institute, Dr. Edmonds of Pacific Northwest National Laboratory, and Professor Jacoby of MIT. 57-713 99-44 actual and potential GDP loss. expenditures for purchases of carbon emission permits, change in carbon intensity from the respective reference case, and change in fossil fuel consumption. Tables 30 and 31 provide comparisons of the results for 2010 and 2020. Further details are provided in Appendix C. For the WEFA study, comparisons are provided only with the 1990-7% case. For DRI comparisons are provided only for a trading case (Case 2). WEFA does not believe that sinks, offsets, or trading will be agreed upon and implemented before the target period of 2008 to 2012, nor by 2020. As noted earlier in the report, EIA does not have the capability to analyze international trading and thus is unable to provide a most likely estimate of the impacts of the international trading provisions of the Kyoto Protocol, or of sinks and offsets, on the level of the energy-related carbon reductions required to meet the 1990-7% reduction in greenhouse gases. EIA's 1990+9% and 1990+14% cases are used in Table 31, because the carbon emissions levels of those cases were most closely aligned with the other studies presented. Some of the major factors that result in differences in the projected carbon prices and costs to achieve the 1990-7% carbon reduction level are: • Relative differences in reference GDP and carbon emissions growth rates through 2020. For example, if the GDP or carbon emissions growth rate in a given reference case is lower than that in EIA's reference case, a smaller carbon reduction will be needed, and it will generally be easier to achieve the emissions target. If the reference GDP growth or carbon emissions growth is higher than in EIA's reference case, the carbon price and GDP impacts relative to those projected by EIA in this study will generally be higher. Most of the major differences among the analyses are attributable to differences in the reference case projections. • Differences in assumptions about the potential for economical life extension or refurbishment of existing nuclear power plants beyond their normal licensing period. If, for example, no existing nuclear plants were retired by 2020, about 40 million metric tons of carbon emissions would be avoided from the combustion of fossil fuel used in plants to replace them. • The amount of knowledge about future events assumed for decisionmakers. For example, models that assume that decisionmakers have perfect knowledge about future prices, demands, or policies could underestimate compliance costs, because al. future events would be anticipated with certainty and responded to at minimum cost. Analyses that assume that all decisionmakers are myopic will tend to overstate transition costs. • The amount of lead time decisionmakers are assumed to have to adjust to the Kyoto Protocal For example, if a model starts to begin the adjust ment process in 1985, 1990 or 1995, it could underes timate the costs of complying with the Kyoto Protocol, because it has more time to adjust. Modes that wait until the last moment to begin the adjust ments could overstate adjustment costs. • The level of aggregation in the model for technolo gies and goods. A model that deals only with aggre gate products such as oil, gas, or coal without the benefit of an explicit technology representation may not capture important variables that can signif cantly affect energy efficiency and intensity or the changing mix of industries that may result from compliance efforts. • The amount of focus on the transition process and the associated costs. For example, a model that assumes that all capital and labor can be immediate ly switched from one use to another cannot capture the short-term or medium-term impacts of complying with the Kyoto Protocol, because those costs are not reflected in the model. • The assumed speed and extent of changes that consumers can make in energy consumption or demand for energy services in response to chang ing prices (price elasticities of demand). Higher assumed elasticities make it easier to achieve the carbon target through demand reductions. Lower elas ticities make it more difficult. Among the studies compared in Table 30, the projected carbon prices in 2010 fall into three groups. MIT, EPRI CRA, and WEFA project prices in the range of $265 (WEFA) to $295 (CRA) per metric ton of carbon. PNNL projects carbon prices of about $221 per metric ton. ELA projects carbon prices of $348 per metric ton. Further contributing to the low carbon price projection is the amount of lead time consumers have to respond, as well as differences in the reference case economic growth rates. The PNNL and EIA reference case GDP projections are very similar. However, PNNL's end-use representation does not explicitly represent technologies, and PNNL's assumed consumer responsiveness to prices (prompting lower energy service demand) and interfuel substitution potential appear to be substantially higher in the medium term (through 2010) than the implicit elasticities in EIA's explicit representation of technologies and consumer choices. The PNNL model begins solving in 1985 in 5-year increments. The PNNL reference case is calibrated to AEO98. In the PNNL policy runs, the carbon policy was phased in over a 10-year period beginning in 2000. Consequently, policy adjustments begin in 2001, consumers and producers begin to anticipate the Kyoto Protocol in that year, making the appropriate adjustments. In the PNNL analysis, electricity demand grows by 0.4 percent annually in the reference case between 2010 and 2020. This is a significant departure from the annual growth rate of more than 2 percent in recent years. Most electricity demand projections have annual growth in excess of 0.9 percent between 2010 and 2020, as compared with PNNL's 0.4 percent." Offsetting these factors are factors that tend to overstate cost. For example, in the PNNL analysis, primary renewable use for generation changes only slightly from the reference case in 2020, even with a carbon price of $286 per metric ton. 96 The group of models projecting costs between $265 per metric ton and $295 per metric ton in 2010 for the 19907% case include transitional processes and costs-either in the macroeconomy or in the energy system-through a detailed representation of the cost, performance, and market adoption of technologies. This group includes the CRA model. Through 2010, CRA projects that, in the reference case, U.S. GDP will grow by $270 billion more than projected in most of the other studies compared. The higher growth rate of GDP normally makes the reduction in emissions harder and more costly to the U.S. economy. 97 If differences in the reference cases were the only factor accounting for the different estimates of the costs of complying with the Kyoto Protocol, then CRA's costs would exceed EIA's and WEFA's in 2010; however, large econometric models of the U.S. economy like those of WEFA and DRI tend to focus on the transitional process. including the method of recycling any carbon fees that may be collected by the Federal Government, and unemployment that may be increased as a result of policy implementation. The WEFA, EIA, and DRI analyses assume that labor can be dislocated, whereas most other analyses assume full employment" despite the sudden reduction of energy resources. More aggregated world analyses, including the CRA, PNNL, EPRI, and MIT studies, omit such details, because the inclusion of global regional coverage and trade flows requires simplifications (some important) in the detail with which each region is represented. Model aggregation tends to underestimate the macroeconomic costs; on the other hand, a lack of global coverage (as in the EIA, DRI, and WEFA models) may overstate transition costs, particularly if international trading is implemented efficiently. Also, fossil fuel consumption in 2010 in the CRA analysis is about 6 quadrillion British thermal units (Btu) less than in the EIA reference case, with virtually identical carbon emissions levels, suggesting an accounting difference in emissions coefficients. 92 The PNNL study uses a dynamic-recursive, computable general equilibrium (CGE) model with neoclassical elements. A model is a *general equilibrium" model if it represents all parts of the economy, both energy and non-energy, and all markets clear (supply equals demand at the prices determined). The model is "computable" if a computer is used to solve for the equilibrium; it is "dynamic" if it keeps track of variables over time. A model is "neoclassical" if the model structure assumes that (1) its economic agents have perfect foresight and knowledge of all past, present, and future events, (2) there is perfect and instantaneous ability of capital and labor to move between uses and sectors, and (3) such transitions are costless and instantaneous. 93 The PNNL model (SGM) can be run with either perfect or imperfect foresight. Labor and new capital move freely. 94 A carbon price of $221 per metric ton in 2010 would increase the delivered electricity price by 49 to 69 percent and reduce electricity consumption by 22 percent relative to PNNL's reference case. This implies that, on average, consumers will reduce consumption of electricity by 3.2 to 4.5 percent for every 10-percent increase in the price of electricity. In 2020, a carbon price of $286 per metric ton translates to an electricity price increase of 59 to 66 percent, resulting in a 28-percent reduction in electricity consumption. This implies that consumption will decline by about 4.2 to 4.7 percent for every 10-percent increase in price. (The estimated electricity price changes were derived from comparable EIA cases.) 95 For example, WEFA's annual electricity growth rate is 1.7 percent and EIA's is 0.9 percent. 96 The WEFA, CRA, MIT, and DRI models are econometric, general equilibrium, macroeconomic models. WEFA and DRI model the United States, CRA and MIT model the world. 97 The full employment assumption means that the unemployment rate is unchanged from reference case levels. Table 30. Comparison of Results for Reducing Carbon Emissions to 7 Percent Below 1990 Levels EPRI allows 50 million metric tons for sinks in this case. The percentage represents MIT's upper bound estimate, including some macroeconomic adjustment costs. MIT provided a range from 4.5%) -1.5 percent for change in GDP, to be interpreted as minimum and maximum losses to the economy. For the purposes of this chapter, the owls range is the irreducible economic loss. Because GDP was not provided for the MIT reference case, the reader may assume a central value for GDP $9,400 billion in 2010 and $10,900 in 2020 (1992 dollars). Consequently, the range of losses is $52 billion to $156 billion in 2010 (1996 dolars) The losses in potential GDP for ELA shown in Tables 30 and 31 use two dificent concepts, which give slightly different results. One uses the cont putation of potential GDP that is derived from the DRI model as described in Chapter 8 of this report. The second uses the approximation netol under the carbon reduction versus carbon price curve, also discussed in Chapter 6. The two calculations produce nearly identical results for the 1990 3% case. For the 1990-7% case, the DRI calculation produces a smaller estimate of potential GDP losses. For all other cases, the DR calculat produces a higher estimate of potential GDP losses. Because the projections from analyses other than EIA's were calculated using the approximation method related to the carbon reduction versus carbon price curve, estimates from both the DRI and approximation methods are provided for the EIA study. Only total primary energy was provided. Fossil fuel consumption was derived by subtracting an estimate for nuclear energy and renewable energy ranging from 13 to 17 quadrillion Btu from total primary anargy for 2010. Only total primary energy was provided. Fossil fuel consumption was derived by subtracting an estimate for nuclear energy and renewable enery of 12 to 20 quadrillion Btu from total primary energy for 2020. NA = not available. Sources: EIA: National Energy Modeling System, run FD07BLW.D0803988. WEFA: WEFA, Inc., Global Warming: The High Cost of the Kyoto Protocol, Histor State Impact (Eddystone, PA, 1998), PNNL: E-mail of data from PNNL with explanation of GDP effect received from Ronald Sands of PNNL on August 25, 1908. CRA Paul M. Bernstein, Charles River Associates, e-mail communications, August 24, 1998, EPRI: E-mail provided by R. Richels of EPRI on July 6, 1998, MIT: Facemie Sand July 10, 1998, from Prol. Henry Jacoby, MIT, Cambridge Massachusetts. |