Co2 emission limits: an economic cost analysis for the USA.

CO2 Emission Limits: An Economic Cost Analysis for the USA(*)

This paper provides a cost-benefit analysis of controlling or decreasing CO2 emissions. It uses an analytical framework, called Global 2100, which is designed to evaluated CO2 energy economy interactions and estimate the cost

of a carbon emissions limit. It analyzes three demand parameters (potential GNP growth, elasticity of price induced substitution between capital-labour and energy, and the rate of autonomous energy efficiency improvements) which are crucial to the debate over energy and environmental futures. The paper discusses various energy sources which are either presently in use or will possibly be in use in the future, and analyzes their impact on cost-benefit analyses. Finally, the paper analyzes the results of carbon constraints and suggests that there is need for more research and development on the subject.

INTRODUCTION

Within the scientific community, there is a growing consensus that rising concentrations of certain trace gases in the earth's atmosphere may lead to significant changes in climate. The greenhouse effect has evolved from a purely scientific issue to an important public policy debate. During the 100th Congress (1988-89), more attention was devoted to hearings on the climate than to any other single environmental issue, including acid rain. The result has been a steady flow of legislative proposals to limit emissions of the major greenhouse gases: carbon dioxide (CO2), methane (CH4), nitrous oxide (N2O), and chlorofluorocarbons (CFCs).

Figure 1 shows the estimated current contribution to global warming of the various manmade greenhouse gases. CO2 (believed to be responsible for approximately half the problem) is produced primarily from the burning of fossil fuels. Therefore, the energy sector plays a central role in proposed strategies to delay climate change. Over the next few decades, such strategies typically call for a concerted push toward greater energy efficiency and, to whatever extent is possible, switching away from coal and oil toward natural gas with its lower carbon emissions per unit of energy. For the longer term, proposed strategies tend to emphasize greater dependence on carbon-free alternatives, such as solar (in several different forms) fission and fusion.

Although many of the legislative proposals have set physical targets for the reduction of emissions, little attention has been paid to the costs of meeting these targets. This presents a serious dilemma to policy-makers. Without information on the cost of emissions abatement, it is difficult to assess the feasibility of alternative proposals and to determine which measures are cost-effective. Moreover, a reduction in emissions is not the sole policy response to potential climate change. There is a point at which further reductions could become so expensive that it would be preferable to shift to other options, such as adaptation. Without careful analysis, it is difficult to know where that point might be.

This paper describes Global 2100, an analytical framework for estimating the costs of a carbon emissions limit. The model is designed to evaluate CO2 energy economy interactions. Although the analysis is being implemented on a global scale, this paper will only report upon its initial application. Because of the worldwide scope of the problem, it is clear that a unilateral limit would be futile unless there were parallel abatement actions by other nations.

Sections 2 and 3 contain a description of Global 2100, the rationale for its design, and a discussion of key macroeconomic parameters that affect the demand for energy. The next two sections summarize the cost and performance characteristics of both electric and nonelectric supply technologies. We then present a quantitative assessment, estimating the costs of a carbon limit under five alternative scenarios. Results are also presented on CO2 emissions in the absence of any limits, the size of the carbon tax that would be required to reduce emissions to the target levels, and the benefits of various alternatives for reducing CO2 emissions. In the final section, we conclude with some general observations on the costs of a carbon constraint to the USA.

MODEL STRUCTURE

The name Global 2100 has been adopted in order to emphasize both the global nature of the carbon emissions problem and also the need for a long-term perspective. There are long time lags inherent in the buildup of CO2 and in the transition away from carbon-based fuels. Our model is benchmarked against 1990 base year statistics, and the projections cover 11 ten-year time intervals extending from 2000 through 2100. This is an intertemporal rather than recursive model. It is assumed that producers and consumers will be sufficiently farsighted to anticipate the scarcities of energy, and the environmental restrictions that are likely to develop during the coming decades.

In its present form, Global 2100 is based upon parallel computations for five major geopolitical groupings: the USA, other OECD nations (Western Europe, Canada, Japan, Australia and New Zealand), the USSR and Eastern Europe, China, and ROW (rest of world). Each of these areas is endowed with limited amounts of oil and gas resources and each is a contributor to global carbon emissions. Because each region is likely to pursue its own individual interests rather than the global welfare -- and because there are differences in the relative costs of emission abatement -- it would be desirable to analyze this problem within a computable general equilibrium framework. As an initial step in this direction, we make a series of assumptions on the future path of international crude oil prices -- and also place bounds on the willingness of each region to import or export oil. Moreover, it is assumed that if a carbon emissions quota is assigned to each region through international negotiations, there is no practical way to trade these quota rights. At some point in the future, we hope to adopt a computable general equilibrium framework. A CGE framework would allow us to deal explicitly with issues such as trade in carbon quota rights, trade in carbon-intensive commodities, and the impact of carbon quotas upon the international division of labour.

In undertaking a global analysis, we have avoided the data-intensive approach required for end-use models. Because of the difficulties of gathering a consistent international data set and then arriving at a meaningful summary of results, we have adopted a much more aggregative approach than would be appropriate for analyzing the USA by itself. The categories are consistent with those of the International Energy Workshop, and the projections have been benchmarked against that poll's median results. See Manne and Schrattenholzer (1988).

Within each region, the analysis is based upon ETA-MACRO, a model of two-way linkage between the energy sector and the balance of the economy.(1) This is a merger between ETA (a process analysis for energy technology assessment) together with a macroeconomic growth model providing for substitution between capital, labour and energy inputs. (See Figure 2.) ETA-MACRO is a tool for integrating long-term supply and demand projections. It is designed to compare the options that are realistically available to each region as the world moves away from its present heavy dependence upon oil and gas resources toward a more diversified future energy economy. This type of model may help to promote "second order" agreement. For example, two analysts may disagree on the costs of solar electricity generation, but might agree on a logical framework within which to estimate the impact of these cost estimates. ETA-MACRO allows explicitly for:

* energy-economy interactions: rising energy costs and limited

supplies will prevent the economy from achieving its full

potential GNP growth rate, and this, in turn, will slow down

future capital accumulation;

* cost-effective conservation: rising prices will induce

substitution with capital and labour, thereby reducing energy

demands below the amounts projected from historical trends;

* autonomous conservation: changes in government policy and

in the structure of the economy will help to reduce the

amount of energy required per unit of GNP;

* interfuel substitution: changing relative prices will induce

consumers to replace oil and gas with electricity, e.g. heat

pumps in place of fuel burners; and

* new supply technologies: each has its own difficulties and

uncertainties on dates and rates of introduction.

For each region in parallel, a dynamic nonlinear optimization is employed to simulate either a market or a planned economy.(2) Supplies and demands are equilibrated within each individual time period, but there are "look-ahead" features to allow for interactions between periods. These interactions are particularly important for the depletion of exhaustible resources and for the accumulation of capital over time. Savings and investment decisions are modeled so that consumers will receive equal benefits from an additional dollar's worth of current consumption and a dollar's worth of investment.

In order to focus upon the long-run issues of energy-economy interactions, resource exhaustion and the introduction of new technologies, each region is described in highly aggregative terms. Outside the energy sector, all economic activity is represented in terms of dollars of constant real purchasing power. Within the energy sector, only two end products are distinguished: electricity and nonelectric energy.

Figure 2 provides an overview of the principal static linkages between the sectoral and the macro submodels. Electric and nonelectric energy (denoted, by the symbols E and N respectively) are supplied by the energy sector to the rest of the economy. Like the material balance equations of an input-output model, aggregate economic output (Y) is allocated between interindustry payments for energy costs (EC) and "final demands" for current consumption (C) and investment (I). Thus: (1) Y = C+I+EC.

Each component of equation (1) is measured as an annual flow (measured in trillions of constant dollars). For an economy-wide production function,(3) we assume that gross output (Y) depends upon four inputs: K, L, E, N capital, labour (measured in "efficiency" units that represent the sum of labour force growth and productivity gains), electric and nonelectric energy. To minimize the number of parameters that require either calibration or econometric estimation, the long-run static production function is described by a nested nonlinear form: (2) [Mathematical Expression Omitted] Equation (2) is based upon the following assumptions:

* there are constant returns to scale in terms of these four

inputs;

* there is a unit elasticity of substitution between one pair of

inputs - capital and labour - with [Alpha] being the optimal value

share of capital within this pair;

* there is a unit elasticity of substitution between the other pair

of inputs - electric and nonelectric energy - with [Beta] being the

optimal value share of electricity within this pair;

* there is a constant elasticity of substitution between these two

pairs of inputs - the constant being denoted by [Sigma] (= ESUB);

* the scaling factors a and b are determined so that energy

demands in the base year are consistent with the "reference

price" for nonelectric energy; and

* there are autonomous energy efficiency improvements

(AEEI) that are summarized by growth in the scaling factor

b.

Energy-economy interactions occur in two ways. According to (2), energy is an input to the economy. According to (1), energy costs represent one of the claims upon the economy's output. Tighter environmental standards and/or an increase in energy costs will reduce the net amount of output available for meeting current consumption and investment demands. This is why the potential growth rates of GNP do not uniquely determine the realized rates. Since investment determines the accumulation of capital stocks, a lower rate of current investment will, in turn, reduce the amount of output available for future consumption. Alternative carbon emission scenarios will therefore be measured in terms of their impact upon present and future levels of consumption.

In addition to resource depletion, there are the following intertemporal elements of the model:

Savings and investment decisions are determined so as to maximize

the discounted utility of consumption. For simplicity, the utility

function is logarithmic in form. The utility discount rate is chosen so

that the rate of return on capital will remain constant if the realized

growth rate of the economy coincides with the potential rate

determined by the growth of the labour force.

The rate of depreciation of equipment and structures governs the time

lags of demand in response to changing prices. This is sometimes

termed a "putty-clay" approach. That is, the input-output coefficients

for each successive age cohort of equipment are optimally adjusted

to the future trajectory of prices, but there are no changes possible

in these coefficients subsequent to the installation of a given cohort.

Upper bounds are imposed upon the rates of market penetration for

new supply technologies. Moreover, there are lower bounds to

ensure that older technologies will not be phased out too rapidly.

KEY DEMAND PARAMETERS

Three of the demand parameters (potential GNP growth, ESUB and AEEI) are crucial to the debate over energy and environmental futures. There is no easy way to estimate these coefficients econometrically. The values adopted here have been determined so that our model will track closely with the conventional wisdom expressed, for example, by the median poll responses of the International Energy Workshop.

One key parameter is the rate of growth of the labour force and potential GNP. In parallel with the slowdown of population growth during the 21st century, there will be a diminishing rate of growth of GNP, and hence a slowdown in the demand for energy. It is assumed that the USA's potential annual GNP growth rate will be nearly 3% from 1990 to 2000, and that it will slow down to 1% during the latter half of the 21st century. Even with a stationary population, this would allow for a modest increase in per capita living standards.

ESUB represents the elasticity of price-induced substitution between capital-labour and energy. For a demonstration of the importance of this parameter, see Energy Modeling Forum (1977). Over the long run, there is a good deal of possible substitutability between the inputs of capital, labour and energy. The degree of substitutability will affect the economic losses from energy scarcities and price increases. One example of such a tradeoff would be insulation to replace heating fuels in homes and other structures. A second example would be the increased use of heat exchangers and of cogeneration within industry. In the aggregate, the ease or difficulty of these tradeoffs is summarized by ESUB, here taken to be .40. The higher the value of ESUB, the less expensive it is to decouple energy consumption from GNP growth during a period of rising energy prices. When energy costs are a small fraction of total output, ESUB is approximately equal to the absolute value of the price elasticity of demand.

Finally, there is AEEI, the rate of autonomous (non-price-induced) energy efficiency improvements. In econometric investigations of the post-1947 historical record, there has been no evidence for autonomous time trends of this type. (See Brown and Phillips, 1989; Hogan, 1988; and Jorgenson and Wilcoxen 1989.) Technologically oriented end-use analysts, however, have suggested that non-price efficiency improvements may be induced by changes in government policy, e.g. a mandatory doubling or quadrupling of the average fuel efficiency of automobiles during the course of several decades. (See Goldemberg et al., 1987.) Clearly the AEEI parameter is highly controversial. In order to represent two distinct viewpoints, we begin with a zero value for this parameter; and then explore the implications of a high efficiency scenario.

SUPPLY AND COST ASSUMPTIONS FOR ELECTRICITY GENERATION

Table 1 identifies the alternative sources of electricity supply that are included in Global 2100. The first five technologies represent the sources of electricity that exist within the USA today: hydroelectric and other renewables, gas-, oil- and coal-fired units, and nuclear power plants. The second group of technologies includes the new generation options available for the future. These differ in terms of their projected costs, carbon emission rates and dates of introduction. Table 2 contains a summary of our cost and performance estimates for new electricity supply technologies.

Figure 3 shows the breakdown of electricity generation, by source, for 1985. In that year, natural gas-fired plants produced 12% of the electricity in the USA, and coal-fired units produced 57%. Coal produces almost twice as much carbon per kilowatt-hour as natural gas. If sufficient natural gas were available, a conversion from coal to natural gas would therefore make it possible to achieve a substantial reduction in carbon emissions.

It is expected that new gas-fired capacity for base load electricity will be produced by combustion turbine combined cycle plants. These units have a high thermal efficiency and relatively low costs. If natural gas prices remain at their 1988 levels, this technology would represent an attractive source of electricity. In the absence of large-scale discoveries, however, domestic natural gas resources will gradually become exhausted and fuel prices will rise. For example, in a baseline projection for the Gas Research Institute, Woods (1988) projects a tripling of wellhead prices by 2010. With such an increase, gas-fired electricity would lose its competitive advantage over coal.

Two categories of coal-fired technologies are considered -- those without and those with CO2 emissions control. The first category includes both existing and new pulverized coal technologies. Most of the existing pulverized coal plants do not have flue gas desulphurization units. As a result, they have much lower operating costs than new pulverized coal plants.

CO2 can be separated from either the flue gas of an atmospheric boiler or the fuel gas produced within an integrated gasification-combined cycle (IGCC) plant. According to the studies reviewed by Vejtasa and Schulman (1989), the latter appears to be more cost effective for new power plants. Table 2 contains the cost and performance data for coal gasification technologies with 20% and 93% CO2 recovery.

Separation of CO2 does not solve the problem of permanent disposal. Technically, this gas could be injected into the oceans or into depleted natural gas fields. For purposes of Table 2, disposal costs are based upon compressing the recovered CO2 at the generating station, transporting it by a new pipeline for 100 miles, and then disposing of the gas either in the oceans or in distant natural gas fields via the existing pipeline network. The feasibility of these disposal options is highly speculative. In the following analysis, we calculate the benefits of solving the problem of permanent disposal.

ADV-HC and ADV-LC refer, respectively, to high- and low-cost non-carbon based electricity generating technologies. Although any of a number of technologies could be included in these categories, the cost and performance data contained in Table 2 are based upon specific designs considered in EPRI's Technical Assessment Guide (1989). The representative high cost source is an advanced solar technology with cost and performance characteristics similar to those for concentrator photovoltaic cells. (Alternatively, this might be a biomass-based generating unit or some combination of the two.) The low cost source is an advanced nuclear design with passive safety features. In our judgment, 2010 is the earliest availability date for significant amounts of electricity from these technologies.

NONELECTRIC SUPPLY TECHNOLOGIES

The nonelectric energy supply technologies are listed in Table 3. The individual fuels are ranked in ascending order of their cost per million BTU of nonelectric energy. The least expensive is CLDU. This category accounts for direct uses of coal in industries such as iron and steel, cement, etc. Its growth rate is taken to be only 20% that of the GNP. Next in the "merit order" are domestic oil and gas. These exhaustible resources are available at constant marginal costs, but are subject to upper bounds based on a model of reserves and resource depletion.

For specifying the upper bounds on exhaustible hydrocarbon resources, we draw a sharp distinction between current reserves and the remaining stock of undiscovered resources. Because cost estimation is exceedingly hazardous in this area, we do not attempt to provide an explicit economic rationale through rising marginal cost curves. Instead, a constant ratio model is employed to determine an upper bound on the annual rate of oil and gas production. There is also the possibility of delaying the exploitation of these resources.

Reserves of exhaustible resources are depleted by current production and are augmented by new discoveries. Production is a fixed fraction of reserves (the 1990 production-reserve ratio), and new discoveries are a fixed fraction (5% per year) of the remaining undiscovered resources. When the production-reserve ratio exceeds the resource depletion factor (RDF), it can be shown that the RDF governs the ultimate rate of decline. Figure 4 illustrates the comparison between a 3% versus a 5% value of the RDF. With 3% for this parameter, production drops off even more rapidly during the years through 2030.

Our reserve and resource estimates are taken from the 5th percentile point along the probability distributions, available from the Geological Survey work by Masters et al. (1987). This source provides a modal (i.e., most likely) estimate of resources along with the 5th and 95th percentile. For practical purposes, the 95th percentile point indicates a lower bound on undiscovered resources, and the 5th percentile indicates an upper bound. That is, according to the USGS, there is only a 5% probability that undiscovered conventional resources will exceed the 5th percentile values. For Global 2100, we have adopted the USGS upper bound on natural gas resources. Had our calculations been based upon the modal or the 95th percentile, the prospects for domestic natural gas production would be considerably more pessimistic than the case examined here. According to Figure 4, there is no prospect that conventional domestic gas resources will permit a significant expansion of consumption above its 1990 level. Since the production-reserve-resource ratios for domestic crude oil are similar to those for natural gas, a similar conclusion also holds for this hydrocarbon resource.

Except for economic rents, it is assumed that oil imports net of exports (abbreviated OIL-MX) are more expensive than domestic supplies. International oil prices are projected to rise over time, but are assumed to be independent of the quantities imported by the USA at any one point in time. Typically, this option is also pushed to its upper bound -- a limit of 20 quads based upon national security considerations. In the absence of this bound, the USA would import significantly higher quantities of oil. For modeling purposes, it is assumed that oil imports would be limited either by tariffs or quotas, not by "voluntary export restraints."

According to Table 3, there are two high-cost backstop options -- both available in unlimited quantities: SYNF (synthetic fuels based on coal or shale oil) and NE-BAK (e.g., biomass fuels or hydrogen by electrolysis, using a non-carbon based source of electricity). NE-BAK emits no carbon, but is likely to be more expensive than synthetic fuels based upon coal or shale oil. One or the other of these high-cost technologies will impose an upper bound upon the cost of nonelectric energy -- depending on whether or not there is a carbon constraint.

A CONSTRAINED ENERGY SUPPLY SCENARIO

In this section, we explore the implications of a carbon constraint for the energy sector and for the economy as a whole. Carbon constraints can take a variety of shapes and forms. Legislative proposals have ranged from slowing the future growth rate to reducing CO2 emissions to half their current levels. Because of the wide range of options under consideration, Global 2100 has been designed with a great deal of flexibility regarding the imposition of carbon constraints. Here we illustrate the capabilities of the model by calculating the economic costs associated with just one set of emission reduction targets.

Specifically, we investigate the costs of restricting carbon emissions to 1.37 billion tons (their 1990 rate) through 2000, reducing them gradually to 80% of this level by 2020, and stabilizing them thereafter. Although these targets are not as stringent as those contained in some proposed legislation, they nevertheless represent a substantial reduction in future emissions when compared with a business-as-usual view.

The impacts of a CO2 limit will depend on the technologies and resources available for meeting energy demands as well as on the demands themselves. Table 4 summarizes five energy supply-demand scenarios under which the impacts of this carbon constraint will be analyzed. Scenario I represents the most constrained case -- both from the perspective of supply enhancement and demand conservation. On the supply side, we have excluded the coal technologies with CO2 removal capabilities (COAL-R) and the low cost non-carbon based sources of electricity (ADV-LC). We begin with such a highly constrained supply scenario in order to establish a basis for calculating the benefits of alternative generation options having lower CO2 emissions. These alternatives include advanced nuclear power and coal gasification with CO2 removal capabilities.

On the demand side, the distinguishing characteristic of scenario I is the rate of autonomous energy efficiency improvements. We assume a zero value for the AEEI parameter. As in the case of electricity supplies, we start with the most constrained case and then assess the benefits from measures which reduce CO2 emissions.

A carbon constraint will have both direct and indirect consequences for the economy. Because of the absence of low cost alternatives, the economic impacts will be highest in scenario I. To understand the source of these impacts, and to gain some insight into the workings of the model, it will be useful to explore how the energy sector might evolve over the next century under such a highly constrained scenario.

Figures 5a and 5b show a snapshot of the energy sector at two points in time, 2010 and 2030, under scenario I. In each instance, the optimal combination of supply alternatives is shown -- with and without the carbon constraint. Within the electricity sector (Figure 5a), note the importance of price-induced conservation. A carbon constraint limits the options for electricity generation, raises the price of electricity, and this, in turn, drives down demand.

With a carbon constraint, and in the absence of an economical CO2 removal capability or a low cost carbon-free alternative, the options for meeting electricity demand are severely limited. One alternative is greater reliance on natural gas. Indeed, Figure 5a shows a significant rise in the use of this low carbon fuel in the electric sector. Recall that gas-fired plants produced only 12% of total electricity in 1985. According to Global 2100, the share of gas will rise to 27% by 2010 if we are in a carbon constrained environment.

Increased demands for natural gas from the electric sector will eventually place tremendous pressure on natural gas markets. Our calculations show that by 2010 the price of natural gas will approach the point where high cost renewables (ADV-HC) become an attractive source of electricity. Despite the cost, we see a substantial role for this supply category if carbon emissions are constrained and low cost alternatives are unavailable. In such a world, high cost renewables would, by necessity, become the marginal source of electricity supply.

Without a carbon constraint, the picture is entirely different. According to Figure 5a, coal would once again be the fuel of choice and would supply an increasingly larger share of the load. Coal-fired electricity is plentiful and relatively inexpensive. Although natural gas prices would not rise as rapidly as with a carbon limit, the geological resource constraints, and competing demands from the nonelectric sector, will nevertheless lead to significant price increases. The results of Global 2100 are broadly consistent with the estimates that appear in Woods (1988). In his baseline projection for the Gas Research Institute, he projects a tripling of wellhead prices by 2010.

Now consider the nonelectric side of the overall energy balance (Figure 5b). In the short term (through 2010) a carbon constraint would be felt mainly through its impact on the price of natural gas. High prices will, in turn, lead to lower demand through price-induced conservation.

The impacts of a carbon constraint on the nonelectric sector become more pronounced over time. As indicated by Figure 5b, if CO2 emissions were not a concern, the burden of meeting nonelectric demands would eventually shift to carbon based synthetic fuels. However, such a shift is infeasible if there are stringent limitations on carbon emissions. We would then have to rely upon a non-carbon source, such as hydrogen produced by electrolysis (NE-BAK). According to Table 3, this type of fuel is likely to be considerably more expensive than, for example, gasoline from oil shale or from direct coal liquefaction.

Using Global 2100, we may add together the costs throughout the economic system and calculate the annual losses in consumption due to the carbon constraint. Figure 6 shows the losses, in each time period, for scenario I. The effects of a carbon constraint do not begin to have measurable macroeconomic consequences until 2010. At that point the rise in energy prices begins to have a significant effect upon the share of gross output available for current consumption. By 2030, roughly 5% of total annual macroeconomic consumption is lost as a consequence of the carbon constraint. This percentage remains relatively constant for the remainder of the time horizon. Adding all the years from 1990 through 2100, the present value of these losses would be $ 3.6 trillions, discounting to 1990 at 5% per year.(4)

It is instructive to look at the time path of CO2 emissions in the absence of a carbon constraint. From Figure 7 we see that there could be a sixfold increase between 1990 and 2100. This represents an average annual growth rate of about 1.7% per year. Although the increase is large, it should come as no suprise. In the absence of a carbon constraint, carbon based fuels are the most economical source of supply in both the electric and nonelectric sectors.

There are a variety of policy instruments available for reducing emissions to the desired levels. One frequently discussed option is to impose a uniform tax upon those activities responsible for carbon emissions. Figure 8 shows the size of such a tax for scenario I. The tax is relatively low in 2000 ($29 per ton of carbon), then rises sharply as emission limits are tightened. In the absence of low-carbon supply alternatives, consumers are willing to pay a high price to burn carbon based fuels. The tax must be sufficiently high to discourage these demands. By the middle of the 21st century, sufficient additional capacity is available to stabilize the tax at about $250 per ton of carbon. In this scenario, the long-run equilibrium tax is determined by the cost and emission coefficients of the synthetic fuels and nonelectric backstop supply technologies.

REDUCING THE COSTS OF A CARBON CONSTRAINT

Of our cases, Scenario I is the most constrained -- from the perspective of supply enhancement and demand conservation. Experience has shown that energy forecasting, even over a few decades, is a highly inexact art. At best, one can ask a series of "what if questions" in the hope of gaining some insights into the relative attractiveness of various means of reducing CO2 emissions. It is in this spirit that we have examined several alternatives -- first individually and then in combination.

In scenario II, we calculate the benefits which might accrue from the successful development of a cost-effective means of CO2 removal, and disposal of the gas in a manner that prevents it from reaching the atmosphere. Recall that the cost and performance data contained in Table 2 are for coal gasification with 20% and 93% carbon removal.

Scenario III explores the benefits from a low cost non-carbon based source of electricity. Here our cost and performance data are based on those for an advanced nuclear technology with passive safety features. See EPRI Technical Assessment Guide (1989).

Figure 9 shows these benefits in terms of reductions in the costs of a carbon constraint. For developing a CO2 removal and disposal capability with the assumed characteristics, the discounted benefits are about $0.6 trillions (the difference in consumption losses between scenarios I and II). The discounted benefits from a low cost, non-carbon based source of electricity are $1.2 trillions. The benefits from these two technologies would not be additive. To a certain extent, the two technologies would be substitutes for each other. However, because of constraints on the rate that any single technology can be introduced, there are benefits from deploying both.

On the demand side, a major issue is the rate of autonomous energy efficiency improvements. Up to this point, we have assumed that the AEEI is zero. That is, there are no energy efficiency improvements except those that are price-induced. For scenario IV -- the high efficiency case -- we assume an AEEI of 1.0% per year. By comparison with scenario I, energy demands in 2050 would be nearly halved. Figure 9 shows that this huge reduction in energy requirements would significantly lower the cost of a carbon constraint. The total discounted consumption losses would drop to $1.8 trillions. By definition, no consumption losses are imputed to autonomous energy efficiency improvements. It is highly controversial whether such rapid improvements are indeed achievable. Through scenario analysis, we have sought to bound the broad range of viewpoints on this issue.

The rightmost bar on Figure 9 shows the costs of a carbon limit in the best of all worlds defined by these five scenarios. That is, both COAL-R and ADV-LC are available, and autonomous (non-price) demand reductions occur at the rate of 1.0% per year. In this case, the discounted costs of the carbon limit fall to $0.8 trillions. This would be only 22% of the losses incurred under the constraints of scenario I.

Finally, we compare the emissions for all five scenarios -- both with and without a carbon limit (Figure 10). When carbon emissions are unconstrained, the profiles for scenarios I and II are identical. There would be no purpose in deploying the carbon removal and disposal technology (COAL-R) if carbon emissions were unconstrained.

Scenario III includes a low cost, non-carbon based source of electricity (ADV-LC). Such a technology would be attractive for economic reasons alone. If it were introduced in 2010, it would assume an increasing share of the electric load thereafter. This automatically leads to a substantial reduction in the rate of growth of carbon emissions.

Scenario IV differs from scenario I only on the demand side. It includes energy efficiency improvements at the annual rate of 1.0%. This leads directly to lower demands for electric and nonelectric energy -- and therefore to lower carbon emissions. When these lower demands are combined with a low cost, non-carbon based source of electricity, carbon emissions are reduced still further (scenario V). Even with the technological advances implied by scenario V, unconstrained carbon emissions would still exceed our carbon limit by a factor of two.

CONCLUSION

In this paper, we have calculated the macroeconomic impacts of limiting CO2 emissions under alternative scenarios. If emission controls are required there will be significant costs, but it is clear that the nation can reduce the size of the ultimate bill through R&D in both the supply and demand sides of the energy sector. According to our calculations, the combination of potential improvements could reduce the costs of a carbon constraint -- perhaps by several trillion dollars.

These savings will not occur unless there are sustained research and development programs on a wide variety of fronts -- both in the public and the private sector. On the supply side, consistent long-term funding is needed to promote non-carbon energy sources such as solar, fission and fusion. R&D on CO2 emissions control could have a substantial payoff -- provided that there is an economical solution to the problem of permanent disposal. There are also gains to be anticipated from more efficient processes for the conversion of conventional fuels into secondary energy forms, such as electricity.

There may also be an enormous potential on the demand side. Following the oil price explosion of 1973, there has been a remarkable improvement in the efficiency of energy utilization. It is unclear how long his trend will continue. Some of this may be the result of an autonomous time trend, but a good deal may be attributed directly to the price mechanism. Research is needed to clarify the role of conservation, and to ensure that cost-effective options are available to the greatest extent possible.

This paper has focused on the costs associated with one set of emission reduction targets. Given the current state of knowledge, it is unclear whether it would be justified to incur these costs. There remain a number of uncertainties in our understanding of the greenhouse effect and in the likely effectiveness of various countermeasures. As observed by Clark (1985), "every responsible scientific assessment of the last several years has noted (if not always emphasized) how thoroughly uncertainties pervade the carbon dioxide question." These uncertainties will not be resolved in the near future.

It could be extremely costly to wait for scientific certainty on the impact of greenhouse gases upon global climate before committing to a vigorous R&D program. New technologies require many years for market penetration. If it turns out that substantial reductions in CO2 emissions are needed, it will be important to have the means available for achieving such reductions in a timely manner. This can only be accomplished through a sustained commitment to R&D. [Figure 1 to 10 Omitted] [Tabular Data 1 to 4 Omitted]

(*)The research reported in this paper was funded by the Electric Power Research Institute (EPRI). The views presented here are solely those of the authors, and do not necessarily report the views of EPRI or its members. The authors are much indebted to Diane Erdmann and to Lawrence Gallant for research assistance. Helpful comments have been provided by: George Booras, Jae Edmonds, George Hidy, William Hogan, Dale Jorgenson, Stephen Peck, Scott Rogers, Chauncey Starr, Stanely Vejtasa, Gary Vine, Robert Williams and two anonymous referees. (1)For a detailed description of ETA-MACRO, see Manne (1981). (2)The model is formulated and solved by means of the GAMS/MINOS system. See Brooke et al. (1988). In a representative example there are approximately 200 constraints, and 400 variables. The solution of two successive cases -- with and without a carbon constraint -- requires five minutes on a 25 Mhz desk-top computer. (3)For the concepts and terminology of macroeconomic production functions and neoclassical growth theory, see Allen (1968). (4)A 5% discount rate is consistent with the numerical assumptions that underlie the economy-wide production function, equation (2). For all cases reported here, we employ 24% as capital's share of the CNP, 2.4 as the initial capital-GNP ratio and 5% as the net annual rate of depreciation of the Capital stock.

Alan S. Manne, Stanford University; Richard G. Richels, Electric Power Research Institute