May 1996
The Emissions Prediction and Policy Analysis (EPPA) model is a component of an Integrated Framework of natural and social science models being developed by the MIT Joint Program on the Science and Policy of Global Change[*]. It is a detailed, global, computable general equilibrium (CGE) model with a long time horizon and regional as well as sectoral detail. The EPPA model can be used to project economic activity, energy use and greenhouse gas (GHG) emissions for each of 12 regions through the year 2100. The model can also be used to simulate different GHG mitigation policy scenarios and to analyze the impacts and consequences of these policies. Within the MIT Integrated Framework, the EPPA projections of anthropogenic GHG emissions are inputs to a coupled chemistry-climate model. It thereby forms the first link in an integrated analysis of global climate change.
The EPPA model is derived from the General Equilibrium Environmental (GREEN) model, which was developed by the Organization for Economic Cooperation and Development (OECD). The evolution of EPPA from GREEN involved adapting the model to the requirements of the Joint Program's integrated analysis. The processes of improving and extending the model are continuing. This report describes the basic structure of the EPPA model, and outlines the way that assumptions about technology and growth are used in defining alternative simulations. A brief discussion of the solution algorithm is included, as well as an introduction to the policy instruments implemented in the model. Simulation results are presented to illustrate the capabilities and behavior of the model, and anticipated future developments of the EPPA model are summarized.
The Emissions Prediction and Policy Analysis (EPPA) model is a component of an Integrated Framework of natural and social science models being developed by the MIT Joint Program on the Science and Policy of Global Change. It is a detailed, global, computable general equilibrium (CGE) model with a long time horizon and regional as well as sectoral detail. The EPPA model can be used to project economic activity, energy use and greenhouse gas (GHG) emissions for each of 12 regions through the year 2100. The model can also be used to simulate different GHG mitigation policy scenarios and to analyze the impacts and consequences of these policies. Within the Joint Program's Integrated Framework, the EPPA projections of anthropogenic GHG emissions are inputs to a coupled chemistry-climate model. It thereby forms the first link in an integrated analysis of global climate change.
The EPPA model is derived from the GeneRal Equilibrium EnviroNmental model (GREEN), which was developed by the OECD (Burniaux et al., 1992).[1] The evolution of EPPA from GREEN can be described by the several steps that have been taken in adapting the model to the requirements of the Joint Program's integrated analysis. For the most part, the benchmark data set of the GREEN model remains unchanged, as do the basic features of the production, consumption and international trade specifications. The main departure from GREEN has been the re-coding of the model in the GAMS language, using a software system designed for general equilibrium problems (Rutherford, 1994). This change was made to facilitate the testing of alternative model features and policies. The first step in implementing EPPA within this new system was to create a "Cousin" version that mimics the original GREEN model. Further development, including a number of changes in the economic structure of the model, have led to EPPA Version 1.6, which is reported here. The processes of improving and extending the model are continuing.
This paper is organized as follows. Section 2 describes the basic structure of the EPPA model, and Section 3 outlines the way that assumptions about technology and growth are used in defining alternative simulations. Section 4 is a brief discussion of the solution algorithm. Section 5 introduces the policy instruments now implemented in the model. Section 6 presents simulation results, to illustrate the capabilities and behavior of the model. Finally, in Section 7, anticipated future developments of the EPPA model are summarized.
Table I. Key dimensions of the EPPA model.
| Production sectors, i Non-Energy 1.Agriculture 2. Energy-intensive industries 3. Auto, truck and air transport Energy, e 4. Crude oil 5. Natural gas 6. Refined oil 7. Coal 8. Electricity, gas and water Future Supply Technology,b 9. Carbon liquids backstop1 10. Carbon-free electric backstop2 |
Consumer sectors, c 1. Food and beverages 2. Fuel and power 3. Transport and communication 4.Other goods and services Primary Factors 1. Labor 2. Capital (by vintage) 3. Sector-specific fixed factors for each fuel 4. Land in agricultural |
| Regions (Abbreviation) 1. United States (USA) 2. Japan (JPN) 3. European Community (EEC) 4. Other OECD3 (OOE) 5. Central and Eastern Europe4 (EET) 6. The former Soviet Union (FSU) 7. Energy-exporting LDCs5 (EEX) 8. China (CHN) 9. India (IND) 10. Dynamic Asian Economies6 (DAE) 11. Brazil (BRA) 12. Rest of the World (ROW) |
Gases (chemical formuli) 1. Carbon Dioxide (CO2) 2. Methane (CH4) 3. Nitrous Oxide (N2O) 4. Chlorofluorocarbons (CFC) 5. Nitrogen Oxides (NO)x 6. Carbon Monoxide (CO) 7. Sulfur Oxides (SOx) |
|
1 Liquid fuel derived from shale 2 Carbon-free electricity derived from advanced nuclear, solar, or wind 3 Australia, Canada, New Zealand, EFTA (excluding Switzerland and Iceland), and Turkey 4 Bulgaria, Czechoslovakia, Hungary, Poland, Romania, and Yugoslavia 5 OPEC countries as well as other oil-exporting, gas-exporting, and coal-exporting countries (see Burniaux et al., 1992) 6Hong Kong, Philippines, Singapore, South Korea, Taiwan, and Thailand
| |
The EPPA model is calibrated with 1985 data. The data set consists of Social Accounting Matrices for each of the 12 regions, and a bilateral trade matrix.[3] As an example, Table 2 shows the SAM for the United States. The units of the SAM tables are 1985 US dollars, based on official exchange rates. The SAM dimensions are consistent with the sectoral break-down of the EPPA model. The data set was collected and compiled by the OECD (1993b).
TABLE 2. Sample SAM (USA, 1985, Millions of Dollars)
The model is myopic. It solves for a sequence of new equilibria for future periods, based on assumptions about exogenous trends in the rate of population and labor productivity growth and technology change, as well as endogenous changes in capital stocks and fixed factor supplies.
There are a set of flow aggregates, with precise definitions indicated by the subscripts above:
Then there are the three primary input factors:
The EPPA model, in a modification of the original GREEN version, permits a choice among ways to represent factor substitutability within production sectors, with either putty-putty, mixed vintage, or fully-vintaged specifications for capital inputs. The "putty" portion of the production structure is a multi-layer CES function calibrated with the 1985 SAM tables for the 12 regions. Eqs. 2.2 to 2.7 are EPPA's CES production functions in the order of top-down nesting, where the rs indicate the CES substitution parameters. The corresponding substitution elasticities are shown in Table A1 of Appendix 2.
[5] The structure is shown in Figure 1.
Figure 1. Production Nesting, Backstop Technology, and Consumption.
Eq. 2.1 is the top-layer of the production function. It states that sectoral output X is a linear function of intermediate inputs Xa and an interim quantity variable, Z, comprised of labor, energy, capital and fixed factor,
[Eq. 2.1] 
[Eq. 2.2] 
[Eq. 2.3] 
[Eq. 2.4]
[Eq. 2.5] Eq. 2.1 is the top-layer of the production function. It states that sectoral output X is a linear function of intermediate inputs Xa and an interim quantity variable, Z, comprised of labor, energy, capital and fixed factor,
[Eq. 2.1] [Eq. 2.2] [Eq. 2.3] [Eq. 2.4]
[Eq. 2.5]
[Eq. 2.6]
[Eq. 2.7]
Equilibrium factor demands, by sector and region, and prices for factors and sectoral outputs, can be derived in the usual manner from the primary and dual solutions of the cost minimization problem with the above technology.[6]
The three capital vintaging specifications in EPPA differ in the extent to which the factor proportions can change after the initial investment in a capital good is made. In all three cases, current technology and input prices determine the factor proportions of new investment. In the putty-putty case, there is no distinction between old and new capital: the inherited capital stock adjusts its factor proportions and technical efficiency to changes in relative prices and technology as fully as does new capital. By comparison, in both the vintaged cases, the factor proportions and technical efficiency of the inherited capital stock are frozen for all periods subsequent to its initial creation. In effect, each vintage of the surviving capital stock is Leontief, with technical coefficients determined initially by the CES functions described above and the input prices and level of technology in the initial period. In addition to the differing assumptions about the malleability of the inherited capital stock, the fully vintaged specification makes a different assumption about the survival of the capital stock. In the mixed vintage specification, "old" capital is one composite aggregate for which the technical coefficients are the average of the Leontief coefficients of the constituent vintages weighted by the depreciated quantity of each vintage. In the fully vintaged specification, each depreciated vintage is identified separately for five succeeding periods and then eliminated. As a result of this truncation, the quantity of "old" capital is somewhat less than in either the putty-putty or mixed vintage specifications.
[Eq. 2.8]
Emissions are denoted EE, with the particular gas and sector of origin indicated by subscript indices. Recall that i indicates a production sector, e is a subset of i indicating an energy sectors, and that b is a backstop sector. These sectors are listed in Table 1. To these we add two new indices:
G = {g}, indexing types of gases, also listed in Table 1, and
S = {s}, indexing subsectors of origin of emissions.
With these definitions we can describe how emissions are calculated. Sample results for CO2 emissions are shown in Figures 2 and 3; corresponding results for the trace gases are shown in Figure 4.
Carbon Dioxide. Aside from emissions from deforestation, all CO2 emissions
are ascribed to the region where they are generated. For each region, emissions are calculated as
[Eq.2.9]
Emissions from deforestation are exogenously determined, and are added to the emissions driven by the activity levels generated by the EPPA model. In the results shown below, these emissions are assumed constant at estimated 1985 values from 1985 to 2000, and thereafter to decline linearly to zero in 2050.
Methane. Total methane emissions are the sum of the emissions from a number of sources. First are emissions from agricultural activities, which are driven by activities in the aggregate agricultural sector:
i = agriculture, [Eq. 2.10] 
e = coal, natural gas. [Eq. 2.11] Nitrous Oxide. Emissions of N2O result from a number of agricultural, industrial, and energy-sector activities, and from biomass burning (this last source also treated separately below). The estimates of emissions from fertilizer use and nitric acid production are of the form of Eq. 2.10, with appropriate values of ag,s. Adipic acid production is the same, only related to the Xdi for i = Other Industries and Services. Emissions from fossil fuel combustion takes the form of Eq. 2.11 for e = refined oil, with an appropriate value of ag,e applied. Emissions from the extension of cultivated land are based on a simple growth rate applied to emissions as estimated in 1985. Parameter values are given in Table A4 of Appendix 2.
Nitrogen Oxides. The main anthropogenic source of NOx is fossil fuel combustion from stationary and mobile sources, with additional emissions from petroleum refining, cement manufacture, nitric acid production, and biomass burning. At present, the prediction of NOx emissions from all sources but biomass burning is modeled by an equation of the form of Eq. 2.10, with Xdi from Energy Intensive Industries, and appropriate values of ag,s.
Chloroflurocarbons. Chloroflurocarbons are assumed to follow the phase-out schedule of the London Amendments to the Montreal Protocol.
Carbon Monoxide and Sulfur Oxides. Anthropogenic sources of CO include fossil fuel combustion, the production of steel, pig iron, and ammonia, and the burning of refuse and agricultural waste, sewage, and biomass. The sources of SOx include coal and petroleum use, smelting of non-ferrous ores, production of sulfuric acid, pulp and paper, incineration of wastes, and biomass burning. (As with other gases, emissions from biomass burning are treated separately.) For each region, emissions of these two gases are modeled by two equations each, one representing energy use and the other covering industrial activity.
For CO, the emissions from fossil fuel use are projected as resulting from the use of coal and refined oil:
[Eq. 2.12]
[Eq 2.13]
Biomass Burning. Biomass burning encompasses a variety of activities, which can be roughly classified into the five categories: (1) the burning of the world's forests for land clearing and agricultural use (including both shifting agriculture and permanent agriculture), (2) the annual burning of savannas to control pests, weeds, and brush accumulation, (3) prescribed burning in temperate and boreal forests, (4) the annual burning of agricultural stubble and waste after harvest, and (5) the burning of wood for energy production. These processes emit CO2 and the trace gases CH4, N2O, NOx, SOx, and CO.
In the case of CO2, annual burning of savannas, brushland, and agricultural wastes does not contribute net emissions into the atmosphere, since the CO2 released is re-absorbed by re-growth the following year. Only deforestation with burning of the wood represents a significant net CO2 emission into the atmosphere, as long-term stores of carbon are released with no possibility for re-absorption in the short term. For all other greenhouse-relevant gases, however, all biomass burning yields net releases since they are not re-absorbed by plant growth. With the exception of prescribed fires in temperate forests, biomass burning takes place almost exclusively in the tropical and subtropical regions of the world. Since EPPA does not aggregate into tropical and other regions, we compute emissions from biomass burning as a global value. Agricultural sources of trace gases (category 4 above) are estimated using a relation of the form of Eq. 2.10, with Xdi from aggregated global agriculture, and appropriate values of ag,s. Other sources (categories 1, 2, 3 and 5 above) are assumed to follow the same pattern as the exogenous assumption for CO2 from deforestation.
Annual burning of savannas for control of pests and brush accumulation, prescribed forest fires, and burning due to shifting agriculture is treated as constant over time at their 1985 levels. Shifting agriculture is not expected to grow much, but permanent conversion of forests to land for agriculture or pasture is growing. The harvesting of wood for energy also contributes to deforestation. Emissions from permanent deforestation and fuelwood harvesting grow at the rate of deforestation, which is exogenous as noted above. Burning of agricultural wastes is assumed to increase as agricultural production increases, and is modeled by a relation like Eq. 2.10, with appropriate parameters.
Distribution by Latitude. Emissions of all gases are needed by 7.8 degree latitude band, for inclusion in the MIT coupled chemistry-climate model, because the atmospheric chemistry is influenced by the geographical distribution of emissions as well as their global amount. Global emissions of the long-lived gases (CO2, N2O, CH4 and the CFCs) are distributed by a function that remains constant over time. Emissions of the short-lived gases (CO, NOx, SOx) are distributed within each region according to population distribution, and the latitudinal totals are simply the sum of the 12 regions. Thus with differential growth and emissions among the 12 regions, the latitudinal distribution shifts over time.
[Eq. 2.14]
[Eq. 2.15]
[Eq. 2.16]
[Eq. 2.17]The demand for production goods by the investment sector is defined in the same manner as for consumption:
[Eq. 2.18]
[Eq. 2.19]Government expenditure includes two parts. One is for the direct transfers to the consumer; another is expenditure for inputs required for the production of government services. Government revenue comes from taxes on household and production sectors. The difference between government expenditure and revenue is government savings (Sg). The production function of the government good takes the following form:
[Eq. 2.20]
[Eq. 2.21]Initial current account imbalances among the regions are phased out over time. The absolute values of the imbalances decline gradually during the first four periods, and are eliminated completely in the fifth period. Initial inventory and exogenous investment (i.e., the differences between total investment and household saving in the SAMs) are treated in a similar way. They are reduced gradually in the first four periods, and disappear completely in the fifth period.
Finally, based on the above closure rules, the EPPA model solves for an overall equilibrium in which the following market clearance conditions are met:
[Eq. 2.22]
[Eq. 2.23]
[Eq. 2.24]
[Eq. 2.25]
[Eq. 3.1]
[Eq. 3.2]Labor productivity improvement in region r is expressed as an index over time. It is assumed that all regions start with relatively high, though different, labor productivity growth rates, but converge to some common, lower annual labor productivity growth rate by the year 2100. In the reference runs shown below, this target is 1.0% for OECD regions (USA, JPN, EEC and OOE) and EEX. All other regions converge to 2.0% per annum by the year 2100. Based on the above assumption, Lpr(t), the labor productivity level, is:
[Eq. 3.3]
[Eq. 3.4]Finally, Lr(t), total labor supply is defined as:
[Eq. 3.5]
[Eq. 3.6]
[Eq. 3.7]The treatment of capital formation in EPPA brings neoclassical investment theory into the empirical CGE model environment by making savings a function of the shadow price of capital.[8] This is in contrast to GREEN, which adopts a balanced growth assumption, determines savings exogenously, and adjusts capital productivity (usually negatively) to maintain balanced growth on a pre-determined growth path. In the EPPA model, household savings adjust according to the relative scarcity of capital in the economy. Both r and w are set to 1.0 in the base year. Therefore, if r/w is subsequently greater than one, capital is in short supply, relative to the labor supply in efficiency units, and vice versa. In response to the short supply of capital, household savings are increased as a share of total GDP. Because r/w deviates significantly from 1.0 in a few initial periods, a damping factor is applied to reduce the fluctuation in saving levels. The damping factor in equation 3.7 is set at 0.25 in the current version of the model; and the values of sr are given in Table A9 of Appendix 2.
Within the time horizon of the model, the potential supplies of fixed factors in agriculture, coal and electricity are assumed to have no upper bound. The actual supplies in these sectors are functions of their own prices and previous period's supply level. In fact, FFi,r(t), the fixed factor supply of sector i, region r and period t, is:
[Eq. 3.8] Fixed factor supplies in the oil and gas sectors are exhaustible within the time horizon of the model. In these sectors, the fixed factor supply is proportional to the sectoral and regional resource depletion profiles which are functions of discovered and yet-to-find reserves, an exogenously specified depletion rate and output prices. The determination of resource depletion profiles for oil and gas in the EPPA model is similar to that in GREEN.[9]
[Eq. 3.9]In order to reflect the parallel effect of technological progress on the supply of oil and natural gas, it is also assumed that the fixed factor input requirements for oil and natural gas are reduced over time. The fixed factor improvement rate is set at 1% per year for all regions.
Backstop technologies in the EPPA model have linear production structures with two inputs, labor and capital, which reflect the assumed infinitely reproducible nature of this output. Thus, we have:
[Eq. 3.10]
[Eq. 3.11]
[Eq. 3.12] In the EPPA model, backstop technologies are introduced when
[Eq. 3.13]
[Eq. 3.14]The main advantage of programming in MPSGE/GAMS is that the solution algorithm and the economic model can be separated. This separation makes it possible for users to make changes in model structure, and to introduce new assumptions, without extensive re-programming and debugging. A professional division of labor is allowed, in which the implementation of a robust solution algorithm for CGE models is left to mathematicians and computer scientists, so the modelers can focus on the economic structure of the model.
In the EPPA model, any single solution consists of one assumed equilibrium imposed on the benchmark data for the base year, 1985, and 23 calculated equilibria for five-year intervals through 2100. Each new equilibrium of supply, demand and relative prices is calculated based on exogenous changes in population, labor productivity and the AEEI rate and endogenous changes in saving levels and fixed factor availability. Since the model is myopic, no consideration is given to potential changes in future prices or resource availability in determining each period's equilibrium.
When CO2 reduction policies are imposed, the EPPA model generates a set of new sequential equilibria that are consistent with policy constraints. An entire path with 24 periods generated from policy shocks is a "counter-factual" scenario relative to baseline path.
Both energy taxes and carbon quotas depress the demand for output from the energy sectors. The revenues from taxes and quota allowances are assumed to be transferred to the representative consumer as a lump sum. Since the taxes or quotas distort the equilibria in the economy from the no-policy baseline, the imposition of these policy instruments often reduces GDP. It is possible, however, for regional GDP to increase when the tax or quota has the effect of off-setting the distortion from an existing subsidy. In all the solutions the conventional assumption of CGE modeling is maintained that all productive resources are always fully employed.
In policy comparisons, GDP is often used as an approximate measure of general welfare, and a change in GDP is interpreted as a measure of resources foregone as a result of the policy constraint.[13] In EPPA, we do not use real GDP, strictly defined, as our measure of welfare, but rather real household consumption. We do this to avoid problems involved in the definition of real GDP in this type of model, and the response of this variable to policy perturbations. Real GDP is defined on the input side as the inner product of primary factor prices and quantities. When perturbed by policy, factor quantities adjust very little because factors are fully employed. Therefore, typically real GDP changes little. Household consumption provides a more accurate reflection of welfare. Since consumption consists of four goods in EPPA, we are able to capture the differential effects of policy measures on the composition of household expenditure (for instance, between energy and non-energy intensive goods) as determined by our specification of consumer utility.
Figure 2. Global Carbon Emissions 1985-2100.
As can be seen in Figure 2, the fixity of capital does not matter in the long term because the vintage specification reverts to putty-putty in 2040. The assumptions about backstop technologies do make a discernable difference, however. Given the assumptions incorporated in this EPPA run, the net effect of introducing both carbon-intensive and carbon-free backstop technology is a lower carbon trajectory over the latter half of the 21st century.
The nearer-term effect of the different assumptions about capital fixity is shown by Figure 3, which is an enlargement of Figure 2 for the years 1985-2050. Since there is an assumed tendency toward improved energy efficiency and higher energy prices over time, successive vintages of capital are less energy and carbon-intensive. Global emissions are higher when the energy efficiency of the older vintages is assumed fixed, but by 2025 the initial capital stock has been completely replaced and it is only new technology that matters.
Figure 3. Global Carbon Emissions 1985-2050: The Effects of Vintaging
Figure 4. Global Emissions of CO2 and Trace Gases
In our reference predictions, the paths of emissions with and without the backstop technologies cross over in the 2040s. Initially, the projections of carbon emissions, when backstop technologies are available, are higher than projections without backstop technologies, because the carbon backstop is used earlier than the carbon-free backstop. However, as the carbon-free backstop is deployed, the emissions path for the backstop scenario turns lower by 2050 and is progressively lower as seen in Figure 2.
The main cause of the projected overall increase in global CO2 emissions is growth in regional GDP as shown in Table 3[14], which shows 1985 GDP, average annual rates of growth in subsequent intervals, and 2100 GDP as a multiple of 1985 GDP. The last column shows that regional differences in GDP growth are considerable. With the exception of Japan, the GDP for the OECD regions (the first four rows) increases about five-fold, while regions such as China, India and the Dynamic Asian Economies increase by multiples of 20 to 25 by the year 2100. The differences reflect the effect of assumptions concerning population growth and increases in labor productivity, as well as differences in capital formation.
TABLE 3. Changes in regional GDP.
| 1985 GDP | Growth Rates of Regional GDP | GDP in 2100 | |||||
| Region | (billion 1985 US$) | 1985-2000 | 2000-2025 | 2025-2050 | 2050-2075 | 2075-2100 | (as a multiple of 1985) |
| USA | 3,692.54 | 3.18% | 1.57% | 1.30% | 1.16% | 0.96% | 5.524 |
| JPN | 1,265.99 | 4.03% | 2.18% | 1.70% | 1.29% | 1.03% | 8.398 |
| EEC | 2,227.56 | 3.10% | 1.69% | 1.23% | 0.87% | 0.69% | 4.811 |
| OOE | 787.41 | 3.96% | 1.48% | 1.20% | 1.02% | 0.82% | 5.519 |
| EEX | 1,047.61 | 6.25% | 1.67% | 1.14% | 0.63% | 0.39% | 6.444 |
| CHN | 440.90 | 9.12% | 3.33% | 2.54% | 1.60% | 1.10% | 30.714 |
| FSU | 637.74 | 7.13% | 1.00% | 1.76% | 1.56% | 1.41% | 11.620 |
| IND | 181.79 | 7.04% | 3.65% | 2.66% | 1.89% | 1.44% | 29.901 |
| EET | 237.15 | 6.41% | 2.00% | 1.88% | 1.44% | 1.25% | 12.934 |
| DAE | 261.78 | 6.59% | 4.32% | 2.67% | 1.43% | 0.79% | 25.183 |
| BRA | 186.31 | 4.19% | 4.48% | 2.57% | 0.65% | 0.06% | 12.478 |
| ROW | 614.24 | 4.27% | 3.44% | 2.46% | 1.69% | 1.32% | 16.934 |
Figure 5 shows the evolution of CO2 emissions per unit of GDP for these regions. In general, carbon efficiencies increase over time, due primarily to improving energy efficiency but also due to a shift towards less carbon intensive technologies, chiefly electricity. Regions with particularly notable near-term improvements in carbon efficiency are the Former Soviet Union, China, and Eastern Europe, in which the projections reflect the effect of removing energy subsidies and the movement to more market-based economies. Among the regions with less improvement in carbon efficiency (e.g., EEX and OOE) the emissions trajectories are explained largely by their role as producers of the carbon-intensive backstop. The resulting refined oil, when traded, yields its emissions in the destination region, but production of this backstop generates CO2 as well.
Figure 5. Evolution of the Carbon Intensity of Regional Economies
The net effect of differing regional rates of growth in GDP and carbon efficiency result in shifts in the regional shares of CO2 output as illustrated in Figure 6. The principal change in shares between 1985 and 2100 are the reduction in OECD share (-10%) and the FSU and EET
shares (-9%) and the increase in EEX share (+15%). The reason for the large shift of carbon emissions to the EEX is that this region becomes the world's primary producer and exporter of the carbon intensive backstop, based on tar sands and oil shales.
Figure 6 Regional Shares of CO2 Emissions, Reference Case
The shift of energy shares is presented in Figure 7. The primary shifts in projected global energy use result from the increasing shares of the electric and oil backstops to compensate for fuels which are assumed to be depletable, namely, conventional oil and natural gas. The substitution of backstop oil for conventional oil keeps the aggregate oil share (refined plus crude) relatively constant (36% in 1985 vs. 35% in 2100); but carbon-free electricity supplies gain share (13% vs. 34%) primarily at the expense of natural gas (21% vs. 4%), which in the current data set is assumed to be severely resource constrained. The coal share remains relatively constant (30% vs. 28%).
Figure 7. Energy Shares in Consumption, Reference Case
Figure 8 compares the reference run (with full vintaging) with the AOSIS Protocol, both with and without backstop technology (squares and diamonds, respectively). The reference runs are in black; the imposition of the AOSIS constraint results in the carbon trajectories shown by the open symbols. The policy clearly reduces global emissions regardless of backstop assumption. Figure 9 shows the evolution of regional shares of global emissions under the AOSIS Protocol, in the same manner as Figure 6 did for the reference run. As would be expected, the AOSIS Protocol reduces the projected OECD share in 2100 from 40% to 17%. Figure 10 shows that the AOSIS policy would increase the fuel share of carbon-free electricity in 2100 by 11% at the expense of oil (-6%) and coal (-5%).
Figure 8. Global Carbon Emissions Under Baseline and AOSIS Cases 1985-2000
Figure 9. Regional Shares of CO2 Emissions, AOSIS Case
Figure 10. Energy Shares in Consumption, AOSIS Case
One of the principal advantages of a regional model, like EPPA, is that it can show the trade effects of regionally specific policies, such as the AOSIS proposal. Two effects are particularly notable: the depressing effects of the CO2 constraint, which forces adjustments to less efficient production and a less desirable composition of output, and the changes in patterns of trade, which generate "carbon leakage." Figure 11 shows the cumulative discounted percentage GDP loss for the twelve regions when compared with the reference case. The GDP losses are greater in the OECD regions, but there is GDP loss in all other regions as a result of the reduced demand in the industrialized world and the consequent effects through international trade. Figure 12 shows the associated percentage change in cumulative carbon emissions by region. The reductions in the OECD regions are significant, and, on the average, emissions increase for the non-OECD regions due to leakage effects. The reduction in cumulative carbon emissions corresponding to the reduction of GDP shown in Figure 11 is shown by the white bars. The black bars depict the substitution effects of the restriction in the OECD regions and the leakage effects in the non-OECD regions, and the net change is the sum of the two effects.
Figure 11. Regional Economic Impacts of AOSIS Protocol (1990-2100, discounted at 5%)
Figure 12. Change in Carbon Emissions Due to Substitution Effects and GDP Loss Resulting form AOSIS Protocol
A further feature of EPPA that is important for policy purposes is the more accurate representation of the costs of the policy due to capital fixity. Figure 13 contrasts the cumulative percentage GDP loss for the OECD when fully flexible capital (putty-putty) and the fully vintaged approach is adopted for the first thirty years of the century and for the entire 21st Century. As shown, the putty-putty assumption always understates cost, but more so in the earlier decades than over the entire period of the projection.
Figure 13. Impact of Capital Fixity Assumptions on Costs to OECD of the AOSIS Protocol
Another area of effort concerns sub-sectors of the energy sector. Of particular importance is the imposition, within the model of natural gas, of more realistic costs of (and constraints on) international shipment, and the elaboration of the cost structure of the backstop technologies. In addition, the production structure within the model is being disaggregated, in order to increase the usefulness of the model for policy assessment, and for the prediction of particular gases. In EPPA Version 1.6, for example, as in the parent GREEN model, transport is contained within Other Industries and Services (see Table 1). Transport is being disaggregated, to allow study of policies directed specifically at this sector (e.g., gasoline taxes and vehicle economy standards). Also, we plan to disaggregate the current Agriculture sector, to identify different types of crops, and livestock activity, in order to improve the analysis of emissions of methane and nitrous oxide.
These efforts are accompanied by a continuing effort to improve the estimates of key input parameters, and to update the underlying social accounting matrices. When these various improvements are brought together into a new version of the EPPA model, a revision of this report will be prepared.
Yc: Consumption of good c
Xai,c : Armington good i demanded by consumer good c
Ec : Energy bundle demanded by consumer good c
Id : Disposable income
Ls : Total labor supply (in efficiency units)
Ks : Total capital supply (capital service)
FFsi: Total fixed factor supply of sector i
Sh: Household saving
Gtr: Government transfers to household
Th: Household income tax
i: Gross investment
Xai,i: Armington good of sector i demanded by investment
EI: Energy bundle demanded by investment
G: Government sector
Xai,g: Armington good of sector i demanded by government
Eg: Energy bundle demanded by government
Lg: Labor demanded by government
Kg: Capital demanded by government
Popr: Population of region r
Plr: Labor productivity of region r
Lr: Labor supply (in efficiency unit) of region r
Xb,i: Output of backstop technology b substituting output of sector i
Xc,i: Output of sector i from conventional technology
rlkef : Substitution elasticity between labor and bundle of capital, fixed factor and energy in CES production function
rkef : Substitution elasticity between energy bundle and bundle of capital, fixed factor in CES production function
rkf : Substitution elasticity between capital and fixed factor in CES production function
re : Substitution elasticity between components in energy bundle in CES production function
ri, 2 : Substitution elasticity between imported and domestic intermediate inputs in CES production function (top layer Armington elasticity)
rj, 1 : Substitution elasticity between imported
intermediate inputs from different regions in CES production function (bottom layer Armington elasticity)
qc : Subsistence consumption of good c
mc : Share of good c in utility function
w: Wage rate
r: Return to capital service
pfi : Return to fixed factor service
g1,r : Initial growth rate of population in region r
g ,r : Asymptotic growth rate of population in region r
g0,r : Initial growth rate of labor productivity in region r
gn,r : Terminal growth rate of labor productivity in region r
ap,r : Auxiliary parameter in determining population growth path in region r
bp,r : Another auxiliary parameter in determining population growth path
al,r : Auxiliary parameter in determining labor productivity growth path in region r
bl,r : Another auxiliary parameter in determining labor productivity growth path
sr : Initial saving rate in region r
dr : Benchmark capital depreciation rate in region r
s: Saving rate adjustment damping factor
af,i : Fixed factor supply elasticity in sector i
ge,j : Annual AEEI rate in sector j
ab,i : Labor share in backstop technology for substituting sector i
bb,i : Capital share in backstop technology for substituting sector i
cb,i : Benchmark unit cost of backstop technology for
substituting sector i
| rlkef | rkef | rkf | re | ri,a2 | r i,a1 | |
| AGRIC | 1.0 | 0.7 | 0.2 | 2 | 2 | 3 |
| COAL | 1.0 | 0.7 | 0.2 | 2 | 3 | 4 |
| OIL | 1.0 | 0.7 | 0.2 | 2 | 3 | 4 |
| GAS | 1.0 | 0.7 | 0.2 | 2 | 3 | 4 |
| REFOIL | 1.0 | 0.7 | 0.2 | 2 | 3 | 4 |
| ELEC | 1.0 | 0.7 | 0.2 | 2 | 0.3 | 0.5 |
| ENERINT | 1.0 | 0.7 | 0.2 | 2 | 2 | 3 |
| OTHERIND | 1.0 | 0.7 | 0.2 | 2 | 2 | 3 |
Table A2. Coefficients of Energy Contents (Exajoule per million 85 US$) (TJ85)
| Coal | Oil | Gas | RefOil | Elec | |
| USA | 0.77392863 | 0.21500844 | 0.35049142 | 0.14487117 | 0.03666328 |
| JPN | 0.18333706 | 0.19020428 | 0.20264218 | 0.11774523 | 0.02878860 |
| EEC | 0.44055759 | 0.20084558 | 0.21657994 | 0.12084787 | 0.03611633 |
| OOE | 0.87697120 | 0.24021348 | 0.36637941 | 0.13963108 | 0.06265089 |
| EEX | 1.09960583 | 0.32586471 | 0.34825090 | 0.13934171 | 0.04533677 |
| CHN | 1.56175400 | 0.21069774 | 0.31063511 | 0.13201714 | 0.13791845 |
| FSU | 1.59455539 | 1.18700000 | 2.36671885 | 1.18697639 | 0.29648470 |
| IND | 1.20249187 | 0.35550614 | 0.55303922 | 0.16871833 | 0.07785827 |
| EET | 1.17001266 | 0.33892263 | 0.64022308 | 0.13810134 | 0.08988690 |
| DAE | 0.46830608 | 0.25073499 | 0.20418787 | 0.09398377 | 0.04111171 |
| BRA | 0.30533072 | 0.27395807 | 0.47145186 | 0.12541675 | 0.08605055 |
| ROW | 1.26048079 | 0.22679826 | 0.14891489 | 0.10603939 | 0.04468788 |
Table A3. Coefficients of Carbon Contents (million ton per exajoule) (ee)
| Coal | 24.686 |
| OIL | 20.730 |
| GAS | 13.473 |
| REFOIL | 20.730 |
| ELEC | 0 |
Table A4. Values of the Parameter ag,s Used in the Trace Gas Estimates
| Gas, g | Source, s | a |
| CH4 | Rice Paddies | 0.3 |
| CH4 | Enteric Fermentation | 0.5 |
| CH4 | Animal Wastes | 0.6 |
| CH4 | Landfills | 0.05 or 0.5a |
| CH4 | Domestic Sewage | 1.0 |
| CH4 | Coal Mining | 1.0 |
| CH4 | Natural Gas | 1.0 |
| N2O | Fertilizer Use | 0.8 |
| N2O | Nitric Acid | 0.8 |
| N2O | Gain of Cultivated Land | 0.03 |
| N2O | Adipic Acid | 1.0 |
| N2O | Fossil Fuel Combustion | 1.0 |
| NOx | Fossil Fuel Combustion | 0.3, 0.5, 1.0 b |
| CO | Fossil Fuel Combustion | 0.15, 0.50c |
| CO | Industrial Sources | 0.5 |
| SOx | Fossil Fuel Combustion | 0.3, 0.5, 1.0b |
| SOx | Industrial Sources | 0.5 |
| all | Agricultural ource Biomass | 0.25 |
|
a 0.05 for OECD, 0.5 for non-OECD. b 0.3 for USA, 0.5 for OECD, 1.0 for non-OECD. c 0.15 for OECD, 0.50 for non-OECD. | ||
Table A5. Base Year Population (millions)
| USA | 239.279 |
| JPN | 120.837 |
| EEC | 321.902 |
| OOE | 118.435 |
| EEX | 655.067 |
| CHN | 1040.264 |
| FSU | 277.537 |
| IND | 750.859 |
| EET | 118.160 |
| DAE | 174.429 |
| BRA | 135.564 |
| ROW | 869.011 |
Table A6. Initial Population Growth Rate (% annum) (g1)
| USA | 0.20722 |
| JPN | 0.03418 |
| EEC | 0.09542 |
| OOE | 0.14010 |
| EEX | 1.54756 |
| CHN | 0.48222 |
| FSU | 0.28184 |
| IND | 0.95412 |
| EET | 0.24764 |
| DAE | 0.66783 |
| BRA | 0.80859 |
| ROW | 1.87180 |
Table A7. Asymptotic level of population (ginfinity)
| USA | 1.23 |
| JPN | 1.03 |
| EEC | 1.10 |
| OOE | 1.15 |
| EEX | 4.70 |
| CHN | 1.62 |
| FSU | 1.32 |
| IND | 2.60 |
| EET | 1.28 |
| DAE | 1.95 |
| BRA | 2.24 |
| ROW | 6.49 |
Table A8. Initial Labor Productivity Growth Rate (g0)
| PRD% | |
| USA | 2.000 |
| JPN | 2.500 |
| EEC | 2.000 |
| OOE | 2.000 |
| EEX | 1.500 |
| CHN | 6.000 |
| FSU | 3.000 |
| IND | 5.000 |
| EET | 2.500 |
| DAE | 5.500 |
| BRA | 3.000 |
| ROW | 2.100 |
Table A9. Initial Saving Rate
| USA | 0.13 |
| JPN | 0.25 |
| EEC | 0.25 |
| OOE | 0.25 |
| EEX | 0.30 |
| CHN | 0.30 |
| FSU | 0.35 |
| IND | 0.15 |
| EET | 0.30 |
| DAE | 0.25 |
| BRA | 0.15 |
| ROW | 0.15 |
Table A10. Benchmark Cost of Backstop Technologies (Cb,i)
| Electricity | Refined Oil | |
| USA | 1.540 | 1.4 |
| JPN | 1.209 | 0 |
| EEC | 1.517 | 0 |
| OOE | 2.631 | 1.4 |
| EEX | 1.904 | 1.4 |
| CHN | 7.723 | 0 |
| FSU | 12.452 | 0 |
| IND | 4.360 | 0 |
| EET | 3.775 | 0 |
| DAE | 1.727 | 0 |
| BRA | 3.614 | 0 |
| ROW | 1.877 | 0 |
[1] We are greatly indebted to colleagues at the OECD Economics Directorate for their generous assistance in the early stages of our model development, both in providing the original code and in helping us to understand GREEN.
[2] For detailed information about CGE modeling, see Shoven and Whally (1992).
[3] A social accounting matrix for a region is an extended input-out table in value terms. It describes value-added composition of production sectors and the total demand for each production sector, the latter being made up of demands by other production sectors, consumption, government, investment and export.
[4] X and Xd represent the same variable. Xd is used only in the context of Armington specification.
[5] The Allen elasticity of substitution is related to the CES substitution parameter by the equation s = 1 / (1 - r).
[6] For details on how to derive the factor demand functions and equilibrium prices, see Varian (1995), or Mas-Colell et al. (1995).
[7] Disposable income is defined after savings and is fully consumed (cf. Eq. 2.15 supra).
[8] For theoretical background on investment, see O.J. Blanchard and S. Fischer (1989).
[9] See OECD (1993a) for an analytical exposition of resource depletion profiles for oil and natural gas.
[10] The prices of the factors are set equal to unity in the initial period. In the case of conventional production, unit cost is also initially 1.0 which implies that the sum of the technical coefficients is unity. Thereafter, unit cost evolves according to changes in input prices and the technical coefficients. For the backstops, the technical coefficients are fixed and such that the initial unit cost must be greater than 1.0, else the technology would be economically competitive in the first period and not a backstop. For example, if the backstop and the conventional fuel both require an identical quantity of labor per unit of output, the backstop must require more capital. It may become economic later, however, if the price of capital becomes sufficiently cheap relative to the wage rate or the price of fixed factor for the conventional fuel.
[11] A computable general equilibrium (CGE) problem is a specific case of much more broadly defined MCP problem. See Stephen P. Dirkee and Michael C. Ferris, "The PATH Solver for Complementary Problems," in Rutherford (1994).
[12] The unit can be either in heat contents of the product or in quantity units of the model, i.e. 1985 US$.
[13] For these comparisons, we do not attempt to measure the value of the environmental or other benefits obtained by the policy measures.
[14] This table and all subsequent figures assume backstops and full vintaging.
[15] The actual AOSIS (Alliance of Oceanic and Small Island States) proposal states 2005 as the timetable for the 20% reduction below 1990 levels, but we impose the 2010 date to facilitate comparison with results from other models which use a decadal time step.
[16] For simplicity, we omit subscript of region in the list of variables and parameters whenever it does not cause confusions.
| Top of page |
| |