98 The CRA model uses perfect foresight for investment behavior, which may also contribute to underestimating the costs. It assumes that products (like gas and coal) are not perfect substitutes and capital is not perfectly malleable. Further, the demand for energy is only moderately responsive to price changes, compared to the PNNL model. CRA develops its model parameters using the GTAP database from Purdue University and the International Energy Agency (IEA) database.

99 EPRI's MERGE model is an Aggregate Optimization Model and has perfect foresight. The EPRI model is being rebenchmarked to start in 2000 and should result in higher carbon prices and higher GDP losses in 2010 than are shown in their current analysis.

100 For PNNL, since the model begins solving in 1985, policy Instruments could be introduced as early as 1990. For this study, PNNL reports that the policy instruments for the Kyoto Protocol were phased in beginning in 2001.

helps to reduce equipment costs in the later part of the compliance period.

101

The other major area of disagreement among the projections is the impact on actual GDP. In 2010, actual GDP losses relative to each reference case range from -1.0 percent (EPRI) at the low end to about 4.2 percent (EIA) at the high end. Some economists have noted that the total GDP impact on the U.S. economy of regulatory programs such as the Kyoto Protocol are large, and that the true costs typically exceed direct costs by a factor of two to four, particularly in the few years following implementation. CRA projects a 2.1-percent loss in GDP in 2010 and a 2.4-percent loss in GDP in 2020. This contrasts with the EIA projection of a 4.2-percent loss in GDP in 2010 and a 0.8-percent loss in 2020, a trend returning to the reference case GDP. The EIA projected recovery trend is due to declining real prices after 2012, whereas increasing GDP losses for CRA are due to continued increasing delivered energy prices throughout the projection period and the relative high GDP level in the reference case from which the reductions must be made.

Most of the reasons for the differences in carbon prices also contribute to the differences in GDP losses. For example, perfect foresight and long lead times allow the economy to adjust at minimum cost as in the PNNL, EPRI, and CRA models. In the WEFA analysis, lower GDP growth in the early period allows for lower carbon prices and smaller GDP losses relative to the EIA study. CRA's lower carbon price and smaller GDP losses are attributable to four factors: (1) the lack of representation of a revenue recycling mechanism, (2) the high level of aggregation of the U.S. energy-economy. (3) the length of the adjustment period, and (4) the incorporation of international trade flows.

The GDP losses portrayed in the analyses are not based on the same definitions. EIA, DRI, and WEFA report losses in potential GDP102 and full macroeconomic adjustment costs. CRA and EPRI report losses to potential

GDP plus some but not all of the macroeconomic adjustment costs, because the level of aggregation used to represent the U.S. macroeconomy does not permit a full representation of the macroeconomic adjustment costs. PNNL reports only the direct cost of meeting the required commitment level, i.e., losses in potential GDP. The loss in potential GDP can be estimated for all the studies except MIT and can be combined with payments for international permits to develop "irreducible" losses to the economy arising from compliance with the Kyoto Protocol for each of the two cases (no trading and Annex I trading)." Estimates of irreducible losses to GDP in the 1990-7% case in 2010 are remarkably close, ranging from $60 billion for WEFA to about $94 billion for EIA (in 1996 dollars). The range of irreducible losses in 2020 is $75 billion for EIA to $130 billion for WEFA. WEFA projects the largest potential loss in 2020 because it has the highest carbon prices and its reference case projection of GDP in 2020 is one of the two highest.

103

The GDP comparisons imply that there is a great deal of uncertainty about the actual economic losses that could result from adherence to the Kyoto Protocol, with actual economic losses rising to as high as 4.2 percent of reference case GDP in 2010-particularly for analyses that use highly disaggregated representations of the U.S. economy (EIA and WEFA). The difference between actual losses and potential GDP losses represents macroeconomic adjustment costs, which are viewed by economists as theoretically reducible by optimal fiscal and monetary policies. This may be another factor leading to the wide variation in estimates of macroeconomic adjustment costs. Nevertheless, there is considerable agreement on the level of the potential GDP losses.

All the studies are in close agreement on the change in carbon intensity that must occur relative to each reference case. Reductions in carbon intensities are between 24 percent and 29 percent in 2010 and between 32 percent and 39 percent in 2020.

101 Jorgenson and Wilcoxen, "Impact of Environmental Legislation on U.S. Economic Growth and Capital Costs," in U.S. Environmental Policy and Economic Growth: How Do We Fare? (Washington, DC: American Council on Capital Formation, 1992); "Reducing U.S. Carbon Emissions: An Econometric General Equilibrium Assessment," Resource and Energy Economics, Vol. 15 (1993), pp. 7-25; and P.M. Bernstein and W.D. Montgomery, "How Much Could Kyoto Really Cost? A Reconstruction and Reconciliation of Administration Estimates" (Charles River Associates, 1998).

102 The curve shown in Figure 114 in Chapter 6 of this report summarizes the relationship between the level of control and the marginal cost of that level of control. Hence, at each increment of control, the marginal cost is by definition equal to the economic resources that must be forgone in order to achieve the increment in control. It follows, therefore, that the sum of the marginal costs must equal the total cost of the controls that would be internalized in markets. This is the integral of the area under the curve, shown as area A in Figure 114. Conceptually, this is essentially the same effect that is measured by the unavoidable cost in the reduction of potential GDP in the macroeconomic models. As shown in Figure 115, this measure of the unavoidable costs using the results of the NEMS model is nearly identical to the similar estimate from the DRI macroeconomic model.

103 Furthermore, for the balance of total emissions needed to meet the Kyoto targets, permits would be purchased on the International market. If the marginal cost of control in the United States and the International prices of permits are in equilibrium, then the area B in Figure 114 will represent the total payments for permits, and the sum of the two parts will represent the irreducible losses to the economy under that trading regime to meet the Kyoto requirements.

Comparisons of Annex I Trading Cases

Only five analyses-MIT, EPRI, CRA, PNNL, and DRI-provided simulations of the impacts of sinks, offsets and Annex I trading. DRI's Case 2 is compared with the other Annex I trading cases because carbon permits purchased abroad are closest, falling in the range of 147 to 288 million metric tons. 104 Two EIA cases-1990+9% and 1990+14%-are compared with those studies in Table 31, because both of these cases yield carbon emissions in the range of the other studies. Internationally purchased carbon credits in 2020 fall in the range of 111 to 229 million metric tons for all these analyses. EIA's carbon prices in the 1990+9% case is $163 per metric ton105 in 2010 and $141 per metric ton in 2020. The EIA carbon price in the 1990+14% case is $129 per metric ton in 2010 and $123 per metric ton in 2020. MIT provided only carbon prices and a range of GDP losses; thus, further comparisons are not possible.

EIA's purchased carbon credits in 2010 (229 million metric tons) in the 1990+14% case are closest to the projected international purchased credits by EPRI and CRA (229 and 288 million metric tons, respectively). The carbon price projected in these cases ranges from $109 per metric ton for CRA to $129 per metric ton for EIA, a statistically insignificant variation. While there is considerable agreement on the carbon price and credit purchases in these analyses, actual GDP losses projected in EIA's 1990+14% case are more than 200 percent higher than the actual GDP losses projected by EPRI and more than 33 percent higher than CRA's. It is also about 50 percent higher than DRI's.

In the Annex I trading cases, only the DRI and EIA analyses consider how the domestic funds will be recycled back to the economy. EIA assumed that the revenues from domestic sales of carbon emission permits would be recycled back to consumers through a personal income tax rebate, as described in Chapter 6,108 and DRI assumes a return of funds to business. The DRI choice of returning the carbon revenues to business provides a significant boost to business investment in the economy, which implies higher business profits and lower real incomes for consumers in the medium term. According to the DRI analysis, returning carbon revenues to business ultimately would accelerate recovery

and lead to stronger economic growth in the longer term than would recycling the carbon revenues to consumers. The impacts of the two recycling mechanisms account for most of the differences in macroeconomic results between the EIA and DRI analyses.

The DRI approach also phases in the carbon policy over a 10-year period (an approach necessitated by the structure of the DRI energy model), whereas EIA phases in the policy over a 3-year period. This factor adds to the difference between the EIA and DRI analyses of macroeconomic costs. In the DRI study, the 10-year phase-in and the assumption that consumers will anticipate and respond to the Kyoto Protocol early results in a smoother economic transition and tends to give a lower carbon price than analyses with shorter phase-in periods like EIA's.

The estimates of unavoidable (irreducible) losses-
income losses that cannot be recovered-for the U.S.
economy range from $32 billion (DRI Case 2) to about
$62 billion (EIA) in 2010. There are many frictions that
can increase costs above the irreducible minimum.
These include business cycles, international trade and
capital constraints, regulation, use of imperfect instru-
ments instead of auction permits, coal subsidies, CAFE
standards, exemptions, efficiency losses from taxation,
etc.
Various Federal Reserve and Federal Government
policies might mitigate actual GDP losses. There is con-
siderable uncertainty regarding all the above actions.

107

The EPRI analysis, because of its perfect foresight and optimizing framework, yields actual GDP losses that are closest to its estimated unavoidable losses. CRA estimates actual GDP losses that are almost 3 times its unavoidable losses in 2010, and estimated actual GDP losses in 2010 for the DRI and EIA 1990+14% cases are 3 to 4 times the unavoidable losses. Because DRI's and EIA's actual GDP losses are based on a detailed macroeconomic model that has limited foresight, focuses on the transitional process rather than the steady-state condition of the economy, their projected GDP losses are expected to be the largest and perhaps more appropriate in the mid term (through 2010). WEFA and EIA incorporate revenue recycling, while DRI redirects the revenues through higher profits to business.

104 Standard and Poors DRI recently analyzed three cases for the UMWA-BCOA LMPCP Fund. Case 1 assumed that 8 percent of the necessary carbon reduction in 2010 would be accomplished from sinks and offsets, 15 percent from trading, and 77 percent domestically. Case 2 assumed that sinks and offsets would account for 12 percent of the required reduction from baseline in 2010, 30 percent would be purchased from abroad, and 58 percent would be accomplished domestically. Case 3 assumes that sinks and offsets would generate 16 percent of the required reductions from baseline, 55 percent of the reduction would be purchased from abroad, and 29 percent of the reduction to be accomplished within domestic energy markets. Given that the DRI baseline for 1990 carbon emissions is 1,336 million metric tons, the domestic target for Case 1 in 2010 (1,354 million metric tons) is about 1 percent above 1990 levels, Case 2 (1,452 million metric tons) is about 9 percent above 1990 levels, and Case 3 (1,593 million metric tons) is about 19 percent above 1990 levels.

105 For simplicity and ease of expo tion, it is assumed in this chapter that the carbon price, the price at the margin that the United States is willing to pay to reduce carbon emissions, equals the internationally traded permit price.

106 In Chapter 6, EIA also considers a social security tax rebate.

107 Tom Tietenberg, Environmental and Natural Resource Economics, Third Edition (Harper Collins Publishers, 1992).

The DRI and EIA analyses share the same DRI macroeconomic model; however, they differ in the way they represent the energy market. DRI uses a largely econometric approach, with some technology components to simulate equipment turnover. Responses of energy demand to energy prices are approximated through demand elasticities. Elasticity estimates can vary dramatically and are a major factor in determining results.

108

Because DRI and EIA share the same macroeconomic model, the reference case estimates of macroeconomic variables are nearly identical for 2010. The differences in the reference case energy projections are primarily due to differences in fuel prices. By 2020, the differences between the DRI and EIA macroeconomic projections widen as differences in fuel prices widen.

The EIA 1990+9% case reduces more emissions domestically (325 million metric tons) than the 1990+14% case at an average carbon price of $159 per metric ton (peaking at $163 per metric ton) for the 2008-2012 period. The unavoidable losses to the U.S. economy for 2010 are estimated to be slightly ($3 to $6 billion) more than in the 1990+14% case. The actual GDP losses are more than 3.5 times the unavoidable losses in the EIA cases.

The carbon price in the two EIA cases and the MIT trading case declines from 2010 to 2020, unlike the carbon prices in the EPRI, CRA, and DRI analyses that increase over the decade. Most of the reasons for these differences have already been described in the 1990-7% comparison case and will not be repeated here. However, one noteworthy difference remains the availability and cost of Annex I carbon permits and international trade. In the EPRI model, inexpensive permits are presumed to be available from Russia in the early part of the Kyoto Protocol implementation period but are assumed not to be available in the later part of the period. The elimination of the easy Russian permits makes it harder

for the United States to meet its commitments in 2020 through Annex I trading and raises the carbon permit price by 65 percent relative to 2010. The reason for the 60-percent increase in 2020 in the CRA carbon price is related to the differences in the representation of advanced technologies, the level of aggregation of the CRA model as previously discussed, and the absence of easy carbon permits from Russia.

The Administration's estimate of the costs of implement

ing the Kyoto Protocol109 has been developed, in part, by

using the PNNL model. The Administration's analysis does not provide sufficient data to be included in Tables 30 and 31; however, the Administration asserts in Table 4 of the analysis (page 52) that under Annex I trading. the carbon price would be reduced by 72 percent and the resource cost would be decreased by 57 percent relative to a case in which all carbon reductions are achieved domestically. Using Tables 4 and 5 on pages 52 and 53 of the Administration's report on the Kyoto Protocol, the carbon price for the 1990-7% case can be calculated to be $192 per metric ton (in 1996 dollars), and the irreducible economic losses can calculated to be $60 billion. When Annex I trading is assumed, the Administration projects that carbon prices would be reduced to $54 per metric ton, with $26 billion dollars of irreducible losses.110 The relatively lower GDP growth rate from 1995 to 2010 in the Administration's reference case analysis-2.1 percent annually, compared with 2.3 percent in the AEO98 reference case, is a major factor that results in a lower carbon price and lower economic costs needed to achieve a carbon target.

Based on Tables 30 and 31, the following can be summarized:

• There is no clear consensus on how effective Annex I trading will be in reducing carbon prices and the costs to the United States. WEFA believes that Annex I trading will not be effective at all because of

108

Other reference case differences that influence the Kyoto analysis Include: (1) The DRI reference case projects 3.1 quadrillion Btu lower primary energy consumption and 1.8 quadrillion Btu lower fossil fuel consumption in 2010 than does EIA. By 2020, the differences grow to 4.2 quadrillion Btu of primary energy and 2.4 quadrillion Btu of fossil fuel consumption. Associated carbon emissions are also lower. Consequently, it should be less costly for the economy to achieve the same carbon target (1,452 million metric tons) in the DRI analysis than in the EIA analysis (1.461 million metric tons in 1990+9% case), as Table 31 confirms. (2) The DRI reference case projects higher world oll prices, higher delivered coal prices, and lower gas prices than the EIA reference case and greater coal, lower gas, and lower oil consumption than the EIA reference case for 2010 and 2020. The differences in the mix of fuel consumption are related to the differences in fuel prices in the cases. Because the delivered price that consumers react to is the sum of the fuel costs plus the carbon price, when oil and coal prices are higher (without the carbon price), the additional carbon price required to achieve the same delivered coal and petroleum product prices will be lower. Higher reference case prices imply lower required carbon prices to induce an energy demand or mix change. Lower carbon prices usually result in lower economic losses.

109 The Kyoto Protocol and the President's Policies To Address Climate Change: Administration Economic Analysis (Washington, DC, July 1998). 110 According to Table 5, page 53, of the Administration's report, Annex I trading with participation by key developing countries would result in a permit price of $23 per metric ton and irreducible losses of $12 billion. Table 4 on page 52 of the report indicates that the permit price in that case would be reduced by 88 percent and the resource cost would be reduced by 80 percent relative to a "domestic only" case. This means that 12 percent of the carbon price for the domestic only case would be $23, and thus the carbon price in the domestic only case would equal $192 per metric ton. Similarly, 20 percent of the domestic only resource cost would be $12 billion, meaning that the domestic only resource cost would be $60 billion. Using the percentages for Annex I trading in Table 4, the carbon price and the irreducible losses can also be derived for the Annex I trading case.