Global and Planetary
11( 1996) 20 1-22 1
Multisector economic models for analyzing global climate change Adam Rose The Pennsylvania State University, Department of Mineral Economics, University Park, PA 16802, USA Accepted 30 November
1. Introduction Economics plays a major role in global climate change. Economic models can therefore prove invaluable in understanding the potential of human activity to damage the ecosystem and in designing policies to remedy rhe situation. The purpose of this paper is to describe a set of economic tools, referred to as Multisector Economic Models, and to illustrate their usefulness by applying them to important issues in global climate ch.ange research. There are two major reasons we have chosen to focus on multisector economic models. First, economic activity generates the vast majority of greenhouse gases (GHG:;). Hence, we need models that can depict the creanon and emission of these pollutants and their impacts on the natural world. Ideally, models would include feedback effects as well in terms of the reduction in the quantity and quality of resource and environmental stocks. Useful models would also be able ‘to reflect the key role of technology both in generating GHGs and for their mitigation. These models should also be able to incorporate the rapidly changing nature of technology. Second, just as there is call for higher resolution in general circulation models (GCMS) of climate, there is a similar need to improve the resolution of economic models along several lines. We need a delineation of economic sectors, so that we can distinguish the various activities that contribute, typically very differently, to the problem. This is crucial to predicting greenhouse gas emissions over time in 0921~8181/96/$15.00 $3 1996 Elsevier Science B.V. All rights reserved SSDIO921-8181(95)00053-4
the face of structural changes in the economy and to fine-tuning mitigation policies. It is also useful to know the spatial distribution of potential damages (and the benefits of their avoidance), as well as the cost burden of mitigating them. Such information is necessary for individual countries to make enlightened decisions, both with respect to their own selfinterest and the fairness of the outcome with respect to others. Finally, there is the matter of the socioeconomic distribution of gains and losses within a nation. This is necessary to facilitate public participation in decision-making and to avoid inequities at the most fundamental level-individual welfare. We include the following in the category of Multisector Economic Models (MMs): Input-Output (IO), Social Accounting Matrix (SAM), Mathematical Programming @BP), and Computable General Equilibrium (CGE) Models. These are in contrast to two other major types of models that have also been used frequently by economists in climate change research: Optimal Control Theory (see, e.g., Nordhaus, 1993) and Macroeconomic Forecasting Models (see, e.g., Manne and Richels, 1992). In the sections below, we will discuss MM’s in some detail, but let us first note that, as a group, these models have several common characteristics: ’ . Disaggregation of the economy into a number
’There are literally thousands of contributions cite in this paper, but we will confine ourselves sample of seminal, pedagogical, or state-of-the-art
that we could to even just a references.
A. Rose /Global
and Plunetury Chungr I I (1996) 201-221
of separate activities. This can be done for as many as several hundred producing sectors (see, e.g., U.S. Department of Commerce, 1993) and, in some cases, a large number of household units. Thus, MMs bridge the gap between micro level and macro level analyses. They provide significant detail, including, in some cases, individual consumer and producer decision-making (Shoven and Whalley, 1992). . Comprehensive accounting of input flows. Unlike standard neoclassical production functions that focus on only primary factors like labor and capital, MMs include natural resources and various intermediate (processed) inputs generally grouped under the headings of energy and materials (Hudson and Jorgenson, 1975; Rose and Ayres, 1996). . General equilibrium effects. This emanates from the interaction between various sectors of the economy through their purchases and sales of processed inputs, and between producers and households in terms of the flows of goods and primary factors of production in response to prices and factor returns. This economic interdependence means that actions in any sector set in motion a chain reaction of responses, often referred to as ripple, multiplier, or feedback effects (Miller and Blair, 1985; Dixon et al., 1993). . Technological basis. The production activities of the economy reflect the technology in place at a given time. MMs can also readily incorporate engineering data on technological innovation, as well as links between investment and technological change (Carter, 1970; Leontief and Duchin, 1986). . Linkages to the natural world. This takes place via natural resource input flows, as well as the addition of variables representing natural resource stocks (Duchin and Lange, 1994). One can also readily add sector-specific variables to capture byproducts of production and consumption in the form of emission factors or pollution coeficients. Mitigation efforts can also be conceived of as separate economic activities (Leontief, 1970). Finally, ecological processes can be modeled as a set of activities linking inputs to outputs (Isard et al., 1971). Thus, these models serve as a useful framework for organizing and storing data on economics and ecology separately and in conjunction with one another. They can also serve as the basis for more recent efforts at environmental accounting (CBO, 1994).
. Socioeconomic dimensions. Each of the MMs contain the basis for, or can readily be extended to, distinguishing among major institution accounts, especially socioeconomic status (Bulmer-Thomas, 1982; Pyatt and Round, 1985). Because different groups have different behavioral responses to policy, this helps to improve the accuracy of projections, as well as providing input into value judgments regarding the fair sharing of the policy cost burden or its benefits (Rose et al., 1988). Empirical orientation. The class of MMs represents a pragmatic and operational approach to economic analysis. For example, input-output Tables exist for more than 100 countries, most of them utilizing a set of standard accounting procedures referred to as the System of National Accounts. Moreover, reasonably accurate non-survey approaches have been devised to generate I-O models at the subnational level (see Round, 1983; U.S. Forest Service, 1993). . Spatial economic interaction. These models can readily be linked across regions and nations. In addition to showing the interrelationship between countries, this provides us with a comprehensive picture of the role of individual countries within a world system (Leontief et al., 1987). This is crucial since global problems require global solutions. The above can be seen as strengths of Multisector Models. However, we should point out that not all of them are unique. For example, Econometric Forecasting Models are able to incorporate general equilibrium effects, though they cannot be as explicitly traced as in most of the MMs. The trends in these models have also been to enhance the sectoral detail, though this is often based on fully incorporating or conjoining an I-O model to an Econometric Model (see, e.g., Beaumont, 1990). We should also admit to omissions and weaknesses of the class of MMs. These would include: . Lack of standard statistical properties. The preferable way to construct an I-O model is through an exhaustive census, but costs of doing so are prohibitive. The statistical properties of models based on sample surveys have been justifiably criticized (see, e.g., Gerking, 1976). In the case of CGE models, many of the parameters are gleaned from the literature, meaning, for example, that elasticities of substitution may be incorporated into a single model
A. Rose / Global and Planetary Change I I (1996) 201-221
from diverse studies, data sets, and geographic locations. 2 Still, there are other ways of measuring accuracy beyond those typically used in statistical inference, and MMs have held their own in this regard (Jensen, 1980). . Linearity. All Iof the MMs except CGE models are restricted to linear relationships, which are not necessarily consistent with the real world. However, the models can, at some computational cost, be modified to include nonlinearities or piecewise linear approximations. In iaddition, the linearity of some of the models is misunderstood, e.g., the case of I-O models, where the fixed coefficient requirement does not apply to labor and capital (Chenery and Clark, 1959) and where even material (intermediate) inputs are required in only fixed value shares rather than fixed quantities (Klein, 1974). Individual types of multisector models are often viewed as competing, but we believe they are more reasonably viewed a.soverlapping or complementary. With regard to the overlap, all of the other three types of MMs have an I-O Table embedded in their core. Also, most CGE models are based on a Social Accounting Matrix. No single model is superior for all applications, and the choice is more a matter of which model is most appropriate. To a great extent, the choice of models is a matter of the trade-offs between cost and accuracy within the context of data availability. Many economists typically characterize I-O as the most restrictive of the four model types and CGE models as the least restrictive, since the latter inherently allows for nonlinearities and the role of prices. However, there are instances in which Mathematical Programming and CGE models are more restrictive. For example, they assume optimizing behavior and that the economy i.s always in equilibrium, while I-O models do not. Though the focus of this paper will be on economics, I want to emphasize that I have chosen MMs because of their potential usefulness to natural scientists. At the core of each of the models is an I-O Table, which is a tabulation of all the transactions
‘Perhaps the econometrically sity.
only major exception is the U.S. CGE model estimated by Dale Jorgenson of Harvard Univer-
between sectors of the economy. It is thus a transparent snapshot of the workings of an economic system. Standard I-O Tables clearly depict the role of natural resource inputs (e.g., minerals, energy, and water). With some additional effort, environmental inputs and outputs can be added as well. As such, the advantages of these models as an organizing framework extends beyond economics to the natural sciences. Previous applications have extended to modeling small ecosystems themselves (e.g., Isard et al., 1971) and are capable of being further extended to larger ones, including the atmosphere. No pretense is made that the analyses presented below are flawless. First, they are quite facile with respect to the treatment of environmental variables. Second, they are subject to criticism even within the realm of the social sciences. For example, we utilize growth in Gross Domestic Product as our objective function in the China Study in Section III, when of course this is only a crude proxy for the broader goal of economic development. However, I would like to emphasize that this is more a failing of my work than of the general model types presented. Great strides have been made in recent years to incorporate a multitude of objectives in MMs and other models.
analysis of a conservation
egy 2.1. Introduction
Energy conservation is almost universally considered a prime strategy for mitigating greenhouse gases. At present, for example, 97.9% of the CO, emitted from industrial countries and 70.6% emitted from developing countries stems from fossil fuel combustion (World Resources Institute, 1994). With any potential major shift to renewables many years away, outright reduction in the utilization of coal, oil, and natural gas is an obvious strategy. It appears even more appealing when one considers that a good deal of conservation can be attained at a cost-savings when less energy is used outright or at a zero cost, where, for example, energy-saving equipment must be installed. These factors have led to energy conservation being placed in the broad category of “no
and Planetary Change I I (1996) 201-221
regrets” strategies (see, e.g., Cline, 1992). This refers to measures that do not incur added costs even if projected warming trends are not forthcoming. Clearly, production cost-savings and preservation of our energy resources are pluses. However, to date, very few studies have focused on the potential down-side. Clearly there are jobs and profits at stake in the energy industries. Moreover, declines in fossil fuel sectors will lead to declines in output in successive rounds of upstream suppliers (e.g., mining equipment, fuel service companies) as well as some downstream customers (e.g., railroads, electric utilities). It is not clear whether these negative effects will be offset by the increased efficiency of the economy, various factor substitutions, purchasing power improvements for consumers, or any multiplier effects stemming from increased production of energy-saving equipment. In a recent study, two colleagues and I estimated the effects on the U.S. economy and its energy sectors of a conservation strategy to reduce CO, emissions. The analysis illustrates the usefulness of the basic I-O model in analyzing global warming policy. 2.2. Input-Output Analysis
= the sales by sector i to each of the intermediate sectors j.
Three assumptions enable Eq. (1) to be converted into a model capable of analysis and prediction. They are that: (a) each commodity or service is provided by a single production sector, and that there are no joint products; (b) each sector’s inputs bear a direct proportional relationship to that sector’s output; and (c) there are no externalities. Assumption (b) may be written as: Xij = aijXj
Substituting (2) in Eq. (1) yields the basic I-O model: Xi= 5 aijXj+Yi j= I
Xi=Xii +Xi, + ...+Xi. + Yi (i= l...n)
total gross output of sector i, = the autonomous final demand for the products of sector i, =
The endogenous elements of (3) may be rewritten as: ‘ij
the model’s technical coefficients. The balance equation can also be written compactly in matrix notation as: x=AX+y
Input-Output Analysis was developed by Nobellaureate Wassily Leontief (1936; Leontief, 1941). The approach has gone through 50 years of refinement, and is today the most widely-used tool of regional economic impact analysis (see, e.g., Miller and Blair, 1985; Rose and Miemyk, 1989). The basic I-O model can be defined as an operational, static, linear model of purchases and sales between sectors of the economy, based on the technical relations of production. The static, open I-O model is based on a table of transactions between sectors of an economy. This table can be expressed as a system of accounting identities:
We solve for the vector of annual gross output needed to deliver the exogenously given set of final demands as follows: X= (I-A)-‘Y
The (I - A)- ’ matrix is known as the Leontief Inverse. Each element represents the total direct and indirect effect on the gross output of sector i corresponding to a one unit change in final demand for good j. The sum of elements in a given column is known as the output multiplier for that sector. This matrix is assumed to be constant over short periods of time and enables the I-O model to capture the less apparent, but often sizable, ripple effects of a change in exogenous policy decisions or economic conditions, thusly: AX= (I-A)-‘AY
We utilized the 1990 IMPLAN U.S. I-O table (U.S. Forest Service, 1993), which is a non-survey
and Planetary Change 1 I (1996) 201-221
based update of the official U.S. Department of Commerce, 1982 “benchmark” table. The table is in an “industry by industry” format, and we have aggregated it to the 2-digit Bureau of Economic Analysis Industry Classification levels (82 sectors>. There are some relative advantages to the following I-O simulations over more complex modeling approaches. Since the model allows for limited price responses and substjtution possibilities, it might provide a useful estimate of the short-run response, or otherwise as upper bound impacts. Second, since the analysis is performed in a recursive manner, and thus separates the responses to various tactics embodied in the direct mitigat:ion cost function. 2.3. Basic considerations The empirical estimation of abatement strategies for each sector of the economy can be complicated. We utilized a pragmatic approach, based on conservation levels and elasticity measures that capture the essence of most tactics in an overall mitigation strategy. Greenhouse gas mitigation strategies can be depicted by a step function beginning with the lowest cost options. The first step of this cost function would be no regrets (costless or even cost-saving) conservation. This c’ould result from either a technological innovation or the correction of a misallocation (e.g., eliminating energy-wasting practices). Estimates range from Manne and Richels (1992) their interpretation of autonomous energy efficiency improvements (AEEI), on the order of OS-1.0% per year. More optimistic estimates of costless conservation in the literature are in the range of 20% or more overall (e.g., OTA, 1991; NAS, 1990; and Lovins and Lovins, 1991). 3
3We acknowledge that with respect to current conservation options there are costs to: (a) acqking information about conscrvation alternatives and (b) adjusting behavior. Moreover, new technologies have associ,lted development costs. It is not clear whether these are factored into the OTA and NAS estimates. Still, these factors would not preclude some level of “costless” conservation on net. Of course, there are a large number of examples of process integrated technological changes that do lead to cost savings, but they are inchlded in the various conservation teactics modeled in this paper.
Another major form of conservation is price-induced, i.e., a response to changes in relative prices, as would be stimulated by our initial change in fuel prices as a result of a carbon tax or various types of permit regimes. All incur some costs unless the elasticity of substitution is infinite. Basically, we categorize these options under the headings below following Cline ( 1992): IFFS OFS PMS NFFS -
Inter-fossil fuel substitution Other factor substitution Product mix substitution Non-fossil fuel substitution
The final category of mitigation tactics, though limited in the near-term, refers to “end-of-pipe” options. These usually involve additional capital equipment at no cost savings. We note some limitations of this approach, with these shortcomings also being applicable to most CGE and macro-forecasting analyses, since they utilize constant elasticity measures. Implicitly, each tactic would be used in sequence up to the permit price. This means that many tactics would be used up to their limit. For example, there could be complete substitution away from coal in many sectors and substitutions away from fossil fuels in general (assuming this is the less costly of the various other tactics). Of course, there are restrictions on such substitutions, i.e., there are no alternatives for gasoline and electricity in many processes. Moreover, elasticity measures are typically based on the historical range of experience. Our cost function goes beyond that, and, without some explicit constraints, is likely to be overly flexible. Some realism could be incorporated through external estimates of no-regrets conservation (e.g., engineering estimates of conservation practices relating to each fuel by sector of use). Our simulations are based on a policy variable obtained from a three-part model, the first module of which computes the global carbon tax (or equivalently, the equilibrium price per unit of intemationally tradeable CO, emission permits). This is calculated as a tax (or permit price) of $38.35 (see Rose and Stevens, 1993). This international tax is then at the rate confronted by each sector of the U.S. economy. Applying this policy variable to data on fuel
und Planetcrry Change I I (1996) 201-221
adapted from Montgomery (1992), direct fuel increases are as follows:
Fuel Natural gas Oil Coal
physical unit producer price
Cost as % of
2.4. Basic results In examining the conservation tactic, we first specified a case in which the entire 12.8% cutback in CO, emissions is met by conservation. This means a direct reduction in the purchase of each fossil fuel in all intermediate (goods used to produce other goods sectors and final demand sectors (personal consumption, investment, and government spending). Note that the 12.8% level falls about halfway in-between the pessimistic and optimistic conservation potentials noted earlier. The simple I-O multiplier approximations of general equilibrium effects are shown in Table 1, including total gross output and employment effects on individual fuels and on the economy as a whole. Obviously the impacts are uneven across sectors and are most heavily borne by the energy industries. Output and employment reductions in the fossil energy sectors range from a high of 30.6% in oil and gas extraction to a low of 16.0% in refined petroleum. The economy-wide impact is a reduction of only 2.3%. This may seem small at first, but note that in absolute terms it would mean the loss of 2,500,OOO
Table 1 Economic
impacts of the basic conservation Gross output (million $1990)
Employment (person years)
Coal Oil/gas Refmed petroleum Electric utilities Gas utilities
-$5.540 - 81,979 - 35,825 - 36,580 -21,871
- 54,443 - 227,975 -24,916 - 117,671 - 47,273
Economy - wide
Source: Rose et al. (1994).
26.4 30.6 16.0 16.7 20.9
26.4 30.6 16.0 16.7 20.9
jobs. The results indicate that second-order impacts will achieve some additional mitigation; so that there is actually a need for direct conservation cutbacks of less than 12.8% because of the approximately 10% energy use reductions achieved through direct and indirect output declines. The direct reductions necessary may only be about two-thirds of those originally stipulated and dramatize the potential importance of general equilibrium feedbacks in a more comprehensive analysis. Note also that the results mean the energy sectors can achieve emission reductions without having to undertake any specific mitigation other than their own output reduction. The 12.8% CO, reduction target could be implemented by mandating conservation as opposed to a market incentive approach using taxes or permits. The command and control approach would leave prices unchanged (assuming again that conservation up to the target level is costless). The incentive approach, on the other hand, would raise the price of basic fuels and essentially inspire what amounts to price-induced conservation. The analysis of our first tactic corresponds more closely to the command and control case, or to an incentive approach where each sector’s optimal response is a 12.8% direct reduction in energy intensity. Actually, the latter is highly unlikely for two reasons. First, the optimal response will most assuredly be differential levels of conservation across sectors. Second, it will also most definitely call for differential cutbacks in the use of fuels given their different CO, emission rates. Because of the difficulty of modeling such non-linear and pricerelated responses with an I-O model, we do not perform the analysis of this case here. We can, however, draw some conclusions about the results. They would yield lower overall fuel reductions than the uniform conservation case but with a much higher percentage of the reduction borne by the coal industry* The effects of price-induced conservation are analyzed next with our I-O model. We can calculate a set of energy price multipliers of the impacts of a carbon tax or permit policy on energy and all other sectors. These are presented for individual fuels and the economy as a whole in the fist column of Table 2. We then utilize these results to simulate changing product mix as a mitigation tactic. We multiply the price changes by a set of final demand elasticities to
and Planetary Change
Table 2 Economic impacts of the product mix response, year 2000 Sector
Employment Price change Gross output (million $1990) (person years) (percent)
Coal Oil/gas Refined petroleum Electric utilities Gas utilities
92.6 31.6 16.4 8.0 15.2
- $727 -6664 -5621 - 3380 - 1815 -$191,836
-7144 - 18,531 - 3909 - 10,873 - 3923 - 1,476,767
Source: Rose et al. (1994.). a Weighted average.
determine the direct and indirect impacts in various sectors. The results are presented in Table 2 in terms of direct and indirect reductions in output and employment for energy by fossil fuel type and for the economy as a whole. They show this tactic can have a significant impact across many sectors (not just energy), The economy-wide price increase of 2.9% results in a modest reduction in energy industry output, in comparison to Table 1, but a sizable economy-wide outpm reduction of 1.4%. As with our first tactic, it translates into “mitigation” of greenhouse gas emissions via cutbacks in output, and thus reduces the number or intensity of other tactics that need to be utilized. In essence, if one is modeling a market incentive approach, the results presented in Tables 1 and 2 should be combined, yielding an economy-wide gross output reduction of 3.7% and energy industry output reductions as high as 33% (oil and gas extraction). Again, we note that the emission reductions stemming from the general equilibrium responses are much larger than those initially required so that the (combined results would only be relevant if a :ery myopic policy target were to be implemented. 2.5. Sensitivity analysis An issue that arises in the previous analysis is the manner in which conservation is costless. If the fuel
’This analysis still only analyzes the response to price rises in fmal goods and services. Adjustments to increases in input prices would also set off a set of responses in the production of intermediate goods and thus further accentuate these results.
input per unit of output decreases by 12.8%, might this set into motion production or price responses that could offset in part, in full, or more than fully, the direct and indirect reductions stemming from energy conservation itself. There are several possibilities. First, if the conservation is totally due to inefficiencies, then there is a cost reduction equal to 12.8% of fuel input costs, which are in the range of l.O-5.0% in most sectors. This could be passed on entirely as a price reduction, though this is unlikely. Any reduction in the direct price of each good, however, would potentially reduce the price of all goods in which it is a direct and indirect input, thereby generating price multiplier effects analogous to those discussed above. This would increase the purchasing power of consumers and lead to direct and indirect increases in output and employment. Second, higher profits that would also be associated with the cost reduction could be paid out as higher returns to capital. A similar outcome would take place if conservation stems from substitution of capital for energy. Either of these would result in a direct and indirect stimulus to the economy. Third, the cost decrease would raise the productivity of labor and warrant higher returns to this primary factor. A similar positive stimulus would ensue from the substitution of labor for energy. Fourth, if the conservation is brought about through the substitution of energy-saving equipment, then there is an expansionary stimulus due to the manufacturing of that equipment. There may also be expansions due to increases in the demand for other direct inputs (the general class of materials) substituted for energy. To examine these possibilities, we ran a set of sensitivity tests. The simulations were run by assigning some proportion of the cost savings to each of the four alternatives. For example, since some portion of profits is subject to corporate income taxes, another portion is retained earnings, and because a portion of what is paid out as dividends is saved, we only used 50% of the cost-savings as the direct stimulus in each sector attributable to the direct capital offset. In the case of labor, we examined an upper-bound estimate of 100% of the cost savings and a 75% level, the latter reflecting individual income taxes and savings leakages. For the equipment case, we assigned the entire cost savings level
208 Table 3 Output impacts of conservation Sector Coal Oil/gas Refined petroleum Electric utilities Gas utilities Economy-wide
$ 124 1870 1895 2013 975 $110,011
and Planetary Change II (1996) 201-221
year 2000 a
$ 188 2814 2851 3029 1466
$250 3763 3813 4051 1961
$298 3026 2711 3894 1658 $261,275
$ 152 2647 3342 2307 1204 $165,250
Source: Rose et al. (1994). a In millions of 1990 dollars.
Table 4 Output impacts of the conservation Sector
Coal Gil/gas Refined petroleum Electric utilities Gas utilities Economy-wide
response net of respending
-$5416 -80,109 - 33,930 - 34,567 - 20,896 - $210,165
- $5352 -79,165 - 32,974 - 33.55 1 - 20,405 - $154,658
year 2000 a
-$5290 - 78,216 - 32,012 - 32,529 - 19,910
- $5242 - 78,953 -33,114 - 32,686 -20,213
Price reduction -
$5388 79,332 32,483 34,273 20,667
Source: Rose et al. (1994). a In millions of 1990 dollars.
to the combustion sector as a proxy for all possible energy-saving equipment (and material substitution). In the price case, we simply reduce sectoral prices by 12.8% of total direct fossil fuel costs. The output impacts of conservation cost-savings respending alternatives is presented in Table 3. They range from $110 billion for the capital spending alternative to $261 billion for energy-saving combustion equipment. Still, all five alternatives fall short of offsetting the negative effects on output for the economy in general and the energy sectors in particular (see Table 4). The variation in the energy sector offsets is less than 5% between alternatives. There are two reasons the offsets are smaller than the $320 billion direct conservation output reduction. First, three of the offset alternatives suffer from initial leakages for a variety of reasons (retained earnings, taxes, and savings). Second, the multiplier impacts of energy sector production would appear to be greater than the offsets. Direct comparisons of
these multiplier impacts can be made with the labor (100%) and combustion equipment alternatives. ’ The overall outcome of our experiment depends on which respending alternatives are operative. Of course, the most likely response will be a mix of alternatives. We simulated two alternatives. The first gives each of the four responses equal weight (we utilized only one of the labor alternatives-the 75% respending version). The result is a $145 billion decrease in total gross output over the whole economy in the year 2000, still a 1.04% decrease over
’ We converted the combustion equipment expenditures into annualized equivalents, and thus our simulation results represent an average of impacts over the life of the equipment. To the extent that the equipment will have to be manufactured and put into place in the early stages of the policy time horizon, this offset alternative is actually likely to follow a cyclical path, significantly larger in the first few years and somewhat smaller in the remaining years.
A. Rose / Global and Planetary Change 1 I (1996) 201-221
baseline. The second sensitivity test assigns differential weights based on the judgment of the author. For each of the energy sectors, the results are hardly different than the first test. We should note that the above analysis did not explicitly take the Ibenefits of global warming policies into account. IElsewhere our analysis indicates that the projected total benefits of CO, mitigation will exceed the costs in the U.S. under most policy alternatives (see Rose and Stevens, 1993). There is considerable uncertainty over the gross benefits, but it is still unfortunate that national accounts omit such environmental factors. The conservation strategy could very well lead to an increase in overall augmented GNP, especially if the conservation efforts were part of a more general eco-restructuring process (see, e.g., Ayres and Simonis, 1994). Still, these additional considerations are unlikely to offset the negative impacts o:n the energy industries. Similar results have been obtained when a CGE model has been applied to this issue (see Rose and Lin, 1995).
3. Multisector programming warming policy in China
As the world’s largest producer and consumer of coal, China is an obvious candidate to take a lead role in reducing greenhouse gas (GHG) emissions. However, with a per capita income of less than $400, China is also one OFthe 25 poorest countries in the world. Some analysts suggest that curbing GHG emissions will be very costly. Thus, the question arises as to the equity, if not the feasibility, of having the Chinese initiate major mitigation measures (Rose and Stevens, 1993). Other analysts suggest that efforts to cope with global warming open up a realm of opportunities to improve energy and economic efficiency simultaneously (Cline, 1992). In collaboration with three of my graduate students, I undertook a study to analyze the impacts of greenhouse gas mitigation policies on the growth of the Chinese economy over the next 30 years (see Rose et al., 1996). Our main focus was the reduction of coal use and its implication for reducing CO,, but we also examined the implications for other fossil
fuels, renewable energy resource, and more ordinary air pollutants. Specifically, we examined partial and general equilibrium economic impacts of achieving a 20% reduction in year 2000 baseline CO, emissions by the year 2025, according to a phased-in compliance schedule. We formulated a Dynamic Linear Programming (DLP) model for our analysis. We chose this model format as $e most appropriate to a centrally planned economy since it calls for an economy-wide optimization of GDP subject to various constraints relating to labor, capital, and energy resources, plus limits on air pollution emissions. The model is further extended to a consistent forecasting version that can incorporate alternative assumptions relating to growth rates in GDP, population, wages, and investment. 3.2. Basic structure of the model Linear Programming is a subset of mathematical optimization, and a special case of the Kuhn-Tucker (saddle-point) Theorem. LP is characterized by problems that one seek to optimize a linear objective function subject to linear constraints. A good way to understand the relationship between LP and I-O is through activity analysis in which linear combinations of inputs are used to produce an output or set of outputs (see, e.g., Dorfman et al., 1958). I-O is a simple case of activity analysis where each good is produced by a single technology and there are no joint outputs. LP is in essence a solution algorithm to an activity analysis problem. When we want to impart some choice into an I-O model, such as a
6China has made great strides in moving toward an overall market economy in recent years. Currently, it is a mixed system with public and private enterprises (sometimes in the same industry) operating side by side. While the private enterprises are not state managed directly, the central planning apparatus also refers to broader economic policy-making, which definitely affects them in major ways. The LP model is still very much applicable whether the environmental policies simulated am implemented by decree or by market incentive measures such as a carbon tax. Finally, we note that the model we are using represents the current mix of Chinese enterprises, and that the production relations in that country are likely to change more rapidly over time than in other countries, as generally more efficient private firms displace parastatel corporations.
and Planetary Change 1 I (1996) 201-221
change in the sectoral mix or adoption of new technology to reduce GHG emissions, we use an LP format to optimize this choice. Dynamic LP incorporates time-dependent technologies and changes in capital stocks into a single formulation. Our model represents a modification of that presented in Dervis et al. (1982). Since the model is run for three benchmark years: 2000, 2010, and 2025, it is structured to embody long-run assumptions. Following is an equation-by-equation presentation: 3.2.1. Objective function
5 I;x;S j-
where i? L kj
= = = =
total capital stock (in terms of money value) total labor supply (in terms of money value) capital-output ratio in sector j labor-output ratio in sector j
These equations relate to the stock and availability of primary factors of production. Given the lo-15 year milestone dates during our projection period, we assume perfect capital and labor mobility. 3.2.4. Consumption constraints
where final demand for the products of sector i (ACi+TA,+Ei-Mi) AC, = aggregate consumption of good i (household and institutional consumption) TA, = total accumulation of good i (gross investment plus inventory change) = exports of good i Ei Mi = imports of good i = discount rate P ‘i
3.2.3. Factor constraints
The program maximizes the discounted value of final demand (GDP) over the forecast period. Final demand is composed of the usual aggregate demand categories.
where by b
= rate of economic growth = lower bound = upper bound
This constraint is necessary to ensure that basic human needs are met while other adjustments are made in final demand so as to maximize economic growth. Upper bounds are included to avoid overspecialization or unreasonable shifts in tastes. 3.2.5. Investment, or capital updating, equations
3.2.2. Interindustry constraint X!-
i aijXj’= Yi’ (i= j= 1
K;+’ = K!(l -di) (9)
where aij ‘i ‘i
= the input of good i per unit of output of good j = total gross output of sector i = the autonomous final demand for the products of sector i
This is the I-O balance equation presented in the previous section, which embodies the technological possibilities in the economy.
= depreciation rate
Additional equations for imports, exports, and foreign exchange constraints are specified, but not shown here because of limitations of space. The above is a multiperiod optimizing model that requires a period-by-period specification of the constraint levels for factors (capital and labor) and, for baseline projections, an upper-bound for total output. Those bounds are set by expanding the corresponding baseline figures with growth factors, calculated by adapting Almon (1963) consistent forecasting ap-
Global and Planetary Change I I (1996) 201-221
preach. Almon uses a dynamic input-output model to forecast growth in full-employment economy. No overall optimization of the economy is attempted, however. He assumes that each sector minimizes the costs of capital and labor services of producing output with a Colbb-Douglas production function. Almon’s basic model makes no allowances for factor productivity growth rates, but they can be obtained in a straightforward fashion. The idea is to find the minimum cost of producing a given output level with a Cobb-Douglas production function whose factors are only capital and labor. The first-order conditions are a set of relations between capital and labor for a static picture of the economy. To introduce temporal features, these first-order conditions are converted into growth rates. Optimality relations in levels become optimality conditions for rates of growth. Consistency is provided: to the forecast both dynamically (because growth rates of factors and output are mutually determined within a cost minimizing scheme), and intersectorally (because each sector output is determined by I-O interconnections). We adapted Ahnon’s factor cost minimization, but for the economy as a whole and not on a sector-by-sector basis. The optimization approach translates factor mobility into improved sector allocation and enhances growth. Also, we extend the model to incorporate factor productivity growth (see Rose et al., 1994). 3.3. Basic parameters The CO, emission reduction target used in this analysis is a stabiliz,ation of emissions at 20% of the year 2000 baseline. Because of the projected growth in the Chinese economy, this means a progressively higher overall mitigation level to cap emissions at 80% of their year 2.000 levels over the policy horizon. Moreover, most nations of the world have not agreed to immediate attainment of reduction targets, but to phasing them in according to a compliance schedule (see, e.g., IEA, 1992). Accordingly, we use a phase-in of 25%, 60%, and lOO%, for the years 2000, 2010, and 2025, respectively, that results in effective emission reduction levels of 5.00%, 57.30%, and 84.91%, respectively, for a set of upper bound cases and 5.00%, 40.50%, and 65.31% for lower bound cases (see bellow).
Our empirical model is based on the 1987 Chinese Input-Output Table (PRC, 1990), which we updated and supplemented with data from several other sources (see Rose et al., 1994). The economic parameters used in our simulations were specified in collaboration with Argonne National Laboratory staff and with reference to EWC/ANL (1994). During the past few years, the Chinese economy has undergone phenomenal growth in the range of 8-12%. Based on experiences of other countries, our upper-bound growth rate assumes this pace can continue for a number of years, but that it must decline as the more obvious opportunities for development are exhausted. Our lower-bound estimate reflects conditions under which there might be some domestic strife or some difficulty in the transition of certain sectors to a greater market orientation. The population projection used was based on the upper-bound population projection offered in EWC/ANL (1994). Three major technical aspects of our model should also be noted. The first is emissions of carbon dioxide (CO,) and other air pollutants, such as sulfur oxides (SO,) and methane (CH,). The second is availabilities of energy resources that generate relatively low levels of CO2 or have no direct emissions of air pollutants at all. The third is technological change relating to energy intensity, energy efficiency, and energy substitution. Because of the high degree of uncertainty relating to resource availability and energy technology, we include both a set of upper bound (“optimistic”) and lower bound (“pessimistic”) estimates for these two aspects. Pollution factors express the amount of pollutant emitted per unit of fuel consumed. The pollution factors relevant to the Chinese energy sector came from two sources: those for coal were obtained from ANL (1994) and those for natural gas and petroleum products were taken from EWC/ANL (1994). These pollution factors (tonnes of pollutants per million Yuan of output) are used to calculate emission coefficients by combining them with fuel consumption figures in each sector. Direct and total emission coefficients are used to express the relation between economic activity and air pollution in the Chinese economy. Direct emission coefficients account only for on site, or partial equilibrium, effects and are calculated by finding the amount of each air pollutant emitted by each sector
A. Rose / Global and Planetary Change I1 (1996) 201-221
and then dividing the figure by the corresponding sectoral output. Total emission coefficients account for a combination of direct, indirect, and induced, or general equilibrium, effects and are found by multiplying the direct emission coefficients by the Leontief Inverse Matrix (total input requirements coefficients) of the Chinese economy. These coefficients indicate that many sectors generating rather low levels of CO, directly have a demand for inputs whose production generates a great deal of this greenhouse gas. Electricity supplied by clean fuels is projected to increase from 1990 to 2025. Among the clean fuels considered here are hydroelectric, nuclear, and gasfired power plants. Their share in total electricity generation will be determined by a conjunction of technical and economic factors. The maximum amount of electricity that can be generated by each of these fuels under upper bound assumptions is shown for the years 2000, 2010, and 2025 in Table 5. Technological changes in the energy components of the DLP Model simulations potentially come from three sources summarized in Table 6. For the upper bound case, we assume that there will be no increase in electricity intensity (electricity/GDP ratio). We do, however, envision a full displacement of coal over the planning horizon in all sectors of the Chinese economy except electricity and fabricated metals. Finally, we assume a very high rate of autonomous conservation of 2.5% in the first planning period, tapering off to 1.5% and 1.O% for the subsequent periods.
Table 5 Potential for clean fuels in electricity parameters (in billion 1990 Yuans)
Upper baseline electricity demand Hydro power potential Nuclear power potential Gas-tired electricity potential Total clean fuels as % of baseline Source: Rose et al. (1994).
13.051 0.191 0.268 20.74%
24.446 1.845 71.000 68.51%
45.825 117.556 17.795 532.981 124.108 94.134 64.29% 99.29%
Table 6 Technological change in energy use-upper centage change) 1990-2000 Electricity/GDP Annual average Period total
ratio 0 0
Displacement of coat a Annual average 0.96 Period total 10.00 Autonomous conservation Annual average 2.50 Period total 28.01
Source: Rose et al. (1994). a Applies to all sectors except fabricated metals services. b Applies to all sectors except electricity services.
3.4. Policy simulations Five sets of policy simulations were run, each representing a different abatement strategy and variant of our model: Strategy 1: Change in sectoral mix Strategy 2: Mandated conservation Strategy 3: Interfuel substitution with current technology Strategy 4: Interfuel substitution with technological advances Strategy 5: Combination of strategies The last strategy represents an approximation of the optimal combination of the first four strategies to minimize economic cost (defined in terms of the penalty on economic growth). Note that in all of our analyses, except for Strategies 1 and 5, we have run the following two simulations: (a) with a compliance CO, emission reduction target level constraint; and (b) without an explicit constraint, thereby identifying the maximum attainable emission reduction with that strategy. For Strategy 1, the two cases are: (a) with a CO, emission reduction constraint, plus sectoral constraints on final demand that limit major rearrangements of the economy, and (b) with a CO, emission reduction constraint, but a relaxation of the final demand con-
and Planetary Change 1 I (1996) 201-221
straints. Finally, only one simulation is relevant to Strategy 5. Our results are presented in Table 7. Here, we provide only a brief summary and again refer the reader to Rose et al. (1996). To meet the 5% target CO, reduction for he year 2000, China has several options that do not adversely affect its development plans. In fact, strategies involving conservation and displacement of coal in industrial uses may actually boost economic growth very slightly above baseline projections. By the year 2010, the effective abatement requirement becomes 57.3’0% (upper bound) and 40.50% (lower bound) for the upper and lower bound cases, respectively, of that year’s emissions. This stiffening of the requirement renders most single strategies infeasible, i.e., incapable of achieving the desired emission reduction ‘target on their own. A combination of strategies can attain the target with only a slight decrease in the baseline GNP growth rate in the lower bound ca,se and no decrease in the upper bound one. By the year 20:25, there are fewer options to achieve the target of 84.91% and 65.31% of that year’s baseline CO, emissions. A change in sectoral
Table 7 Summary
bound cases (CO,
0. 1. 2. 3. 4. 5.
Baseline Change in sectoral mix w/ constraint 4b Change in sectoral mix w/o constraint ’ Mandated conservation w/ constraint b Mandated conservation w/o constraint b Interfuel substitution w/ constraint b Interfuel substitution w/o constraint b Technological change w/constraint b Technological change w/o constraint b Combination
mix cannot achieve the results, and autonomous conservation can only achieve reductions of about half the upper bound target and two-thirds of the lower bound one, and only at a very great penalty in both cases. However, inter-fuel substitution and technological change could be combined to attain the targeted reduction of CO, emissions in the upper bound case, and with little economic penalty. Still, the result is highly dependent on our assumption of total displacement of coal in industry and very substantial shifts to nuclear, hydro, and gas-fired electric power generation. Otherwise, the Chinese economy could suffer reductions of economic growth l.O3.0% below baseline. For the years 2000 and 2010, our results do not differ much between upper and lower bound cases, suggesting they are robust over a broad range of assumptions relating to the Chinese economy, its energy reserves, and energy technologies for those years. Unless the CO, reduction measures enacted in China are “sustainable,” i.e., work their way into the very fabric of the Chinese economy and energy system, it could become costly to achieve the reduction requirements in the year 2025 and beyond under either the upper bound or lower bound scenarios.
target: 20% of year 2000 baseline emissions)
GNP growth (%I
CO* reduction (a>
8.50 8.46 n.a. 8.49 8.46 8.51 8.51 8.52 8.59 n.a. ’
n.a. 5.00 n.a. 5.00 25.78 ’ 5.00 9.07 d 5.00 30.59 d n.a. e
7.50 infeasible 2.89 infeasible 7.49 infeasible 7.50 infeasible 7.50 7.50
6.50 infeasible infeasible infeasible 6.49 infeasible 6.50 infeasible 6.51 6.51
n.a. 43.31 d 31.07 d 62.86 d 84.91
57.30 35.23 d 8.62 d 53.19 d 57.30
Source: Rose et al. (19%). a Constraint that sets limits on changes in sectoral output levels. b Constraint that requites CO, emissions be reduced by 5% of year 2000 baseline, 57.3% of year 2010 baseline, and 84.91% of year 2025 baseline. ’ Loosens constraint on sectoral output levels (see text). d Represents maximum achievable reduction. e Abatement target can be achieved by every other strategy; hence, examination of a combination of strategies is redundant.
A. Rose /
Globaland Planetary Change I1 (1996) 201-221
This may have to involve the restructuring of the economy and implementing conservation measures that are no longer costless. In fact, unless major resource discoveries or technological innovations are made, over and above those projected in this report, CO, reduction targets beyond 2025 may not be feasible. This may even be the case for the year 2025 if we have been a bit too optimistic about some of the lower bound set of assumptions. 7 Clearly, there is a need for additional research to improve the accuracy of analyses such as ours. This would include projections of the structural economic changes likely to take place in the Chinese economy with and without tightened environmental regulations. Equally important would be improved estimates of reserves of fuel alternatives to coal (especially renewables) and of technological options that can reduce the demand for fossil fuels or that can render their combustion less injurious to the environment.
of the average size nation. Finally, both economic and ecological data are much more sparse at the regional level. Still, there is a danger in proceeding merely with national level analyses. Both ecological and economic impacts of global warming policy are likely to vary significantly by major region, and the national average belies the extremes (both negative, and, in a few instances, positive). Moreover, new principles of public decision-making, especially those related to the environment, call for greater public participation. These are more effective at the regional and local levels. ’ One of my students recently developed a regional computable general equilibrium (CGE) model (Li, 19941, and we used it to simulate the economic impacts of a carbon tax on the Pennsylvania economy (Li and Rose, 1995). The model advances the state-of-the-art of regional CGE analysis by incorporating special features of economic-energy-environmental policy.
4. Computable carbon tax
4.2. The Pennsylvania
analysis of a
Nearly every major economic analysis of the global warming issue to date has been performed at the national level for several reasons. Negotiations are best undertaken by national governments. Scientific models, primarily general circulation models (GCMS), still have a rather coarse degree of resolution, currently about the level of national boundaries
’We noted earlier that our model lacks a sophisticated investment equation. This is potentially an important limitation and could significantly reduce the accuracy of our results. However, we suggest that this is not a major factor, since our estimate of investment requirements for the CO, mitigation strategies simulated here are very small (see Rose et al., 1996). Although it would seem that investment costs associated with these strategies would be exorbitant, they are not. China is projected to need a great deaf of additional electricity generating capacity under the baseline scenario. Most of our strategies do not increase electricity demand by much, and, in fact, most of them call for decreases. Moreover, there is sufftcient lead time to switch over to cleaner fuels in the process of building new plants rather than retrofitting old ones.
The Pennsylvania CGE model is a long-run, static, single-region model. It is specially designed to incorporate environmental policy variables and their regional economic impacts. This model is based on the following assumptions: 1. The product and factor markets are competitive, i.e., no economic agents are able to exercise monopoly power in the markets. 2. The State of Pennsylvania is modeled as a small, open economy, i.e., it has little or no influence on the supply and demand of goods in the national economy or their prices. 3. Monetary variables do not play an explicit role, i.e., transactions are in real terms. 4. Technological change (both embodied and disembodied), other than that associated with pollution control, is not explicitly taken into account in the period of study.
8Public participation is one of the major underpinnings of the concept of sustainable development (see, e.g., Brundtland et al., 1987). See also the direct application of MMs in Rose et al. (1989).
and Planetary Change II (1996) 201-221
5. Prices are flexible enough to assure the existence of a general equilibrium; producers and consumers can fully adjust without any time lag after external shocks. The major features of this model are its emphasis on production technology and its focus on trade theory. The structure of input demand in this model resembles that of several well-known national and regional CGE models (e.g., the model developed by Robinson et al., 1990, 1993; and the regional model by Kraybill, 1990). Important modifications are made in the specification of the production technology for energy and materials. In the Pennsylvania CGE model, for example, we use a two-tiered, or nested,
Generalized Leontief cost function to describe the production technology. The first tier allows substitutions among capital (K), labor (L), energy (E), and materials (Ml. The second tier specifies interfuel and intermaterial substitutions inside the energy and material aggregates. The overall structure of this model is illustrated by a schematic diagram in Fig. 1. Basically, there are four interrelated components: production, trade, demand, and income. Each component is further decomposed into several parts. The arrows in the diagram show the interlinkages of different parts in the model. When all economic agents interact together and jointly optimize their objective functions, a set
____-_---_--_-___-__--f , Prodution I
I FaCtW Rices
I I i
Payments Ta-Xts Transfers Savings etc.
Fig. 1. Schematic diagram of the Pennsylvania CGE model. Source: Li (1994).
und Planetary Change I I (1996) 201-221
of endogenous equilibrium prices is determined to clear all markets in the economy. The specification of production technology in an empirical analysis requires the selection of a suitable functional form. We rejected Cobb-Douglas, Constant Elasticity of Substitution (CES), and Leontief functions because of their undesirable properties, such as predeterminate restrictions on the elasticities of substitution. The production function used in this study is the two-tiered Generalized Leontief cost functions, one of a broad class of flexible functional forms, which have relatively minimal restrictions on elasticities of substitution and other desirable characteristics (Chambers, 1988). 9 For an open, regional economy like that of Pennsylvania, trade is very important and should be carefully handled conceptually and empirically. In CGE analysis, there are two approaches to modeling the terms of trade in a small, open economy. One method assumes product homogeneity among countries or regions. The other approach specifies product differentiation by country (or region) of origin into the demand structure. Thus, domestic consumers of commodities within one aggregated sector actually consume a mixture of domestically produced and imported goods. The treatment of trade in the Pennsylvania CGE model follows the traditional small, open economy assumption. Industries are distinguished between traded sectors and nontraded sectors. Trade with other regions and foreign countries is modeled as a two-stage decision process. From the small, open economy assumption, the world prices of imports, the prices of imports from other regions, the world prices of exports, and the prices of exports to other
9Theuse of flexible functional forms is not always without any limitations as many advocates originally expected. The separability assumption widely adopted in economic research imposes some restrictions on their flexibility. The other problem lies in the ability of flexible functions to approximate arbitrary technologies globally, because these functions are only second-order numerical or differential approximations at a point. Even worse are their applicabilities to maintain the desirable properties of technology, i.e., a well-behaved situation, over a wide range of observations (Lau, 1986). For example, in their research on the U.S. agricultural production, Baffes and Vasavada (1989) found that TL and NQ cost functions violate concavity conditions.
regions are all exogenously given. On the import side, the consumer, at the first stage, maximizes the utility of consuming composite goods subject to a budget constraint by substituting between foreign goods and domestic goods. At the second stage, a choice is again made between locally produced goods and goods from other regions. At each stage, the Armington assumption is employed and a CES function is used to aggregate imperfect substitutes. On the export side, the producer, at the first stage, minimizes the cost of producing joint products to furnish domestic consumption and foreign demand. Then, at the second stage, the producer again makes supply decisions between local sale and export supply to other regions. The model closure problem in general equilibrium models is actually a synonym for the overidentification problem in a mathematical simultaneous equation system. The tendency of modelers is often to impose as many restrictions as possible in a single model. In most cases, this will lead to an over-determination of the model. To solve the overidentification problem, either one constraint has to be dropped from the model, or one exogenous variable or model parameter must be set endogenously. This process is called the imposition of a closure rule in general equilibrium models. The closure rule adopted here is classical, which is consistent with the long-run nature of the Pennsylvania CGE model. In addition to the closure rule, a set of macro balancing identities is necessary to make the CGE model consistent. These are producer’s equilibrium, goods market clearing, the government budget deficit, the balance of trade, and an investment-savings identity. The problem of interregional factor mobility is critical. Interregional capital movement can be easily modeled by setting net regional savings as endogenous; interregional labor movement, on the other hand, is more difficult for modelers to deal with. The conventional approach to modeling interregional labor mobility is the expression of migration as a function of explanatory variables based on observed data. The treatment of interregional labor mobility in this model, however, takes another approach. In a manner consistent with the Classical closure rule, we assume a fixed labor supply for our initial simula-
A. Rose /Global
and Planetary Change I1 (1996) 201-221
tions. We then relax this assumption in a sensitivity test that calls for any unemployed workers migrating to other states, or any excess demand for labor being completely satisfied by workers immigrating to Pennsylvania from other states. This approach does not model labor mobility in the conventional way but more as a residual.. However, it is an endogenous determination of this important variable. Note that although we have allowed the labor supply to vary, we are not using a pure Keynesian closure rule, which would allow for an underemployment equilibrium. Such an outcome is reasonable for a (closed) national economy in the long run, but it is more typical (and reasonable) in regional modeling to assume that labor migration moves the nonstructural unemployment rate toward zero in the long run for any given region. 4.3. Global warming policy impacts In the current version of the model, carbon taxes are specified exogenously. Carbon tax levels vary due to differences in model structure and policy applications addressed. Except in the often-criticized high carbon tax rate of Global 2100 (see Manne, 19931, other studies suggest a relatively narrow range of tax levels. We utilize the tax values in the works of Jorgenson and Wilcoxen (1992) and Rose and Stevens (1993) for our policy simulations as follows: Case I
Case II Case III
- Stab:ilizing carbon dioxide emissions at year 2000 levels with a carbon tax of $8.55. - Maintaining year 1990 emissions with a carbon tax of $16.96. - 20% reduction of a year 2000 baseline with a c,arbon tax of $38.35.
Performing model simulations requires further transformation of carbon taxes to a set of ad valorem taxes for coal, crude oil, and natural gas. At first, the average heat content of each fossil fuel type in millions of BTUs per physical unit was obtained from the Pennsylvania Coal Data (Pennsylvania Coal Association, 1993). Next, the carbon emission rate measured in kilograms per million BTUs was derived from IEA Coal Research, 1991. These two factors are multiplied together to generate the carbon tax conversion factor for each fossil fuel, which
results in ad valorem taxes of $27.74 per ton of coal, $4.91 per barrel of oil, and $0.63 per thousand cubic feet of gas. Because of the imposed ad valorem taxes on the outputs of coal, crude oil, and natural gas, the first direct effect of all three carbon mitigation strategies appears in the direct price increases of these fossil fuels. The rising prices induce substitution towards other energy sources, as well as substitution between them (inter-fossil fuel substitution), due to the differences in carbon contents and, therefore, the relative strength of tax burdens. More generally, producers will shift away from energy and towards labor, capital, and material inputs. These adjustments inevitably increase the cost of production. On the other hand, the carbon tax revenue is collected by the federal government and is assumed to offset the government budget deficit. As a result, the whole economy is impacted through the linkages among economic agents and supply-demand relationships in all markets. Case I. Stabilizing CO, emissions at year 2000 levels results in a small but significant impact on the Pennsylvania economy (see Table 8). Real Gross Regional Product (GRP) in Pennsylvania falls by 0.26%, the weighted price level increases 0.34%, and total industry output drops by 0.18%. Total employment remains the same since the macro closure rule assumes that labor is inter-sectorally mobile but total labor supply is fixed. The results also indicate that Pennsylvania cuts back total regional exports and
Table 8 Macro economic change)
Emission target Variable
80% of 2000
Carbon tax revenue Real GRP Weighted price Total industry output Total employment Regional exports Regional imports Foreign exports Foreign imports
486.52 - 0.26 0.34 -0.18 _
965.08 - 0.53 0.70 - 0.37 _
2182.24 - 1.20 1.58 -0.84
- 0.29 -0.11 0.14 0.12
-0.51 - 0.22 0.29 0.24
- 1.12 - 0.48 0.64 0.52
Source: Li and Rose (1995).
218 Table 9 Effects of maintaining
year 2000 CO, emissions
and Planetary Change II (1996) 201-221
results of sensitivity
on model assumptions
Base case: fixed labor/ deficit reduction
Sensitivity 1: labor mobility/ deficit reduction
Sensitivity 2: fixed labor/ government expenditures
Sensitivity 3: labor mobility/ government expenditures
Real GRP Price Industry outputs Employment Regional exports Regional imports Foreign exports Foreign imports
- 0.26 0.34 -0.18 _ - 0.29 -0.11 0.14 0.12
- 0.49 0.33 - 0.43 - 0.29 -0.44 - 0.43 -0.14 -0.12
- 0.29 0.35 - 0.24 _ - 0.37 -0.14 0.06 0.04
- 0.53 0.34 - 0.53 - 0.34 - 0.54 -0.51 - 0.28 - 0.24
Source: Li and Rose (1995).
imports decrease by 0.29% and 0.1 I%, respectively. Total foreign imports and eytorts are are increased by a very small percentage. At the industry (or commodity) level the economic impact of a carbon tax varies widely. A significant price change appears in coal, crude oil, refined petroleum, primary iron and steel, electricity, and gas. The complete pass-through of carbon taxes and the prices multiplier effect account for the large price increase for coal and oil. Other commodities using large amounts of fossil fuels in their production process also experience significant price increases. The change in commodity prices leads to a change in commodity demands that, in equilibrium, changes commodity supplies, and thus industry outputs. The pattern of the impacts is roughly inversely correlated to the effect on prices. Case II. The second simulation yields a set of impacts approximately twice as strong as Case I. The overall macro impact on the Pennsylvania economy
is a decrease in real GRP by 0.53%, an increase in price index by 0.70%, and a drop in total industry outputs by 0.37%. Total regional exports and imports decrease by 0.51% and 0.22%, respectively. Total foreign exports and imports increase by 0.29% and 0.24%, respectively. Case III. The final simulation was for the most stringent regulation, and the results are more than four-fold as strong as Case I and more than two-fold as strong as Case II. The macro results show that imposition of a $38.35 carbon tax rate reduces the Pennsylvania real GRP by 1.20% and total industry output by 0.84%. Also, the average commodity price level increases by 1.58%. Total regional exports and imports decrease more significantly compared to previous runs. The most noticeable micro result is the 104.54% increase in the price of coal and the 9.29% drop in its output. With respect to the qualitative and relative patterns of percentage changes compared to the base case, the pattern of the impacts remains the same as the other two cases. 4.4. Sensitivity analysis
“The small economic impact occurs, of course, because the shock is small in the first place. On the other hand, it also indicates the ability of the Pennsylvania economy to adjust to small external stimuli. On the production side, Pennsylvania industries should be able to adopt new technologies in the long-run by moving away from fossil fuel-intensive processes. In addition, there will be inter-fossil fuel substitution (favoring natural gas) and displacement by non-fossil fuels (such as nuclear power). These phenomena are formulated by two-tiered GL functions in the Pennsylvania CGE model. On the consumption side, Pennsylvania consumers will change their consumption patterns towards less expensive commodities and engage in energy conservation.
We also performed sensitivity analyses to test the robustness, and thus indirectly the validity, of our model. First, we tested the sensitivity of the elasticities of substitution in the model. The point predictions of the second scenario in global warming abatement are robust to uncertainty over elasticities. The mean values for major macro variables in the four experiments are of the same order of magnitudes as the reference case.
and Planetary Change I I (1996) 201-221
The sensitivity type of analysis tests the effect of model assumptions on the policy simulation outcomes. First, instead of imposing factor immobility restrictions, we allow for the possibility of interregional labor immigration. Second, instead of assuming carbon tax expenditures offset the budget deficit, we allow for a commensurate increase in government expenditures. The macro results of sensitivity analyses on combinations of differlent model assumptions are presented in columns 2-4 of Table 9 (the base case is presented for purposes of comparison in column 1). Note also that we are performing the analysis in the context of the year 2000 emission level target. When we assume that labor is mobile between regions, then the negative effects on the Pennsylvania economy are actually increased. As shown in column 2 of Table 9, Gross Regional Product declines by 0.49% in comparison to the base case. This means some neighboring states are not hit as hard as Pennsylvania. Some out-migration of productive laborers fron Pennsylvania, in turn, is the primary reason for the fact that Case 2 impacts are more negative than Case 1. However, it is only one reflection of increased competitiveness; note also that regional and foreign trade declines are lower than in the base case. The decrease in imports is almost commensurate with the lower level of industry output in Sensitivity Case 1 relative to the base case. When we assume that carbon tax revenues are used to increase government expenditures, the negative impacts on the Pennsylvania economy increase slightly with respect to the baseline. GRP decreases by 0.29% and totall industry output by 0.24%. Regional exports and imports also decline by more than the reference case (compare entries in columns 1 and 3). The main reason for the dampening effects are the increased competition for investment funds when government expenditures expand and the differential in the stimulus from consumer spending from reduced tax obligations vs. government spending. Finally, we perform a sensitivity analysis of relaxing both key assumptions together. The results, presented in column 4 of Table 9, are closer to the results of relaxing the labor mobility assumption. This is consistent with direct comparisons between columns 2 and 1, and between columns 3 and 1. This means there are nl3 significant synergistic effects
associated with the interaction of these two assumptions.
5. Conclusion In this paper, I have intended to convey the basic tenor and usefulness of several types of multisector economic models (MMs) in addressing global warming issues. These include the broader implications of energy conservation, the long-term consequences of global warming policy on economic growth, and the regional economic impacts of a carbon tax. The analysis has demonstrated the applicability of MMs to industrialized nations, developing countries, and sub-national regions. These applications by no means exhaust the range of topics and issues that can be addressed at the regional, national, and even global levels (see, e.g., Duchin and Lange, 1994). The models illustrated include input-output analysis, mathematical programming, and computable general equilibrium analysis. Space did not allow us to show explicitly the usefulness of social accounting matrices, but the reader should keep in mind that they represent simply a straightforward extension of I-O models and are also the empirical basis upon which CGE models are constructed. SAMs focus more on institutions than producing and consuming sectors and are especially useful in determining the distributional consequences of natural resource and environmental policies (see, e.g., Rose et al., 1988; 1989). It is hoped that the many readers of this journal who work primarily or exclusively in the natural sciences will find these models useful. They can readily accommodate data on natural resources and the environment and can help determine the effects of the interaction between them and the economy. This should provide a common ground for the collaboration by social scientists and natural scientists that is needed to adequately address global climate change issues.
Professor and Head of the Department of Mineral Economics, College of Earth and Mineral Sciences,
A. Rose / Global and Planrtmy
The Pennsylvania State University, University Park, PA. I wish to acknowledge the input of several professional colleagues and graduate students who were co-authors of papers and reports on which this article is based. These include: Brandt Stevens, Ping-Cheng Li, Juan Benavides, Dongsoon Lim, and Oscar Frias. I am also grateful to David Abler and Roy Boyd, who served as the paper’s referees. I am, of course, solely responsible for any errors or omissions.
References Almon, C., 1963. Consistent forecasting in a dynamic multi-sector model. Rev. Econ. Statistics, 65: 148-61. Argonne National Laboratory, 1994. Analysis of Coal Fuel Cycle in China. SFA Pacific, Mountain View, CA. Baffes, J. and Vasavada, U., 1989. On the choice of functional forms in agricultural production analysis. Appl. Econ., 21: 10.53-1061. Brtmdtland, Cl. et al., 1987. Our Common Future. Oxford Univ. Press, World Comm. Environ. Devel., Oxford. Bulmer-Thomas, V., 1982. Input-Output Analysis in Developing Countries. Wiley, New York. Carter, A., 1970. Structural Change in the American Economy. Harvard Univ. Press, Cambridge, MA. Chambers, R.G., 1988. Applied Production Analysis: A Dual Approach. Cambridge Univ. Press, New York, NY. Chenery, H.B. and Clark, P.G., 1959. Interindustry Economics. Wiley, New York. Cline, W., 1992. The Economics of Global Warming. Inst. Int. Econ., Washington, DC. Congressional Budget Office, 1994. Greening the National Accounts. CBO, Washington, DC. Dervis, K., De Melo, J. and Robinson, S., 1982. General Equilibrium Models for Development Policy. Cambridge Univ. Press, Cambridge, MA. Dixon, P. et al., 1993. Notes and Problems in Applied General Equilibrium Analysis. North-Holland, Amsterdam. Dorfman, R., Samuelson, P. and Solow, R., 1958. Linear Programming and Economic Analysis. McGraw-Hill, New York. Duchin, F. and Lange, G.M., 1994. The Future of the Environment: Econological Economics and Technological Change. Oxford Univ. Press, New York. EWC/ANL (East-West Center, Argonne National Laboratory), and the China Consulting Team, coordinated by the SSTC, 1994. National Response Strategy for Global Climate Change. Draft Rep. Asian Devel. Bank, People’s Republic of China. Gerking, S., 1976. Input-Output as a simple econometric model. Rev. Econ. Statist., 58: 274282. Hudson, E.A. and Jorgenson, D., 1975. U.S. energy policy and economic growth, 1975-2000. Bell J. Econ. Manage. Sci., 5: 461-514.
Change I I (1996) 201-221 Isard, W., Choguill, C.L., Kissin, J., Seyfarth, R.H. and Tatlock, R., 1971. Ecologic-Economic Analysis for Regional Planning. Free Press, New York. Jensen, R., 1980. The concept of accuracy in regional input-output models. Int. Reg. Sci. Rev., 5: 139-154. Jorgenson, D.W. and Wilcoxen, P.J., 1992. Reducing U.S. carbon dioxide emissions: the cost of different goals. In: J.R. Moroney (Editor), Energy, Growth, and the Environment: Advances in the Economics of Energy and Natural Resources, Vol. 7. JAI Press, Greenwich, CT, pp. 125-158. Klein, L., 1974. A Textbook of Econometrics. Prentice-Hall, Englewocd Cliffs, NJ, 2nd ed.. Kraybill, D.S., 1990. A computable general equilibrium analysis of regional impacts of macro-shocks in the 1980s. ‘Thesis. Dep. Agric. Econ., Virginia Inst. Technol., VA (unpublished). Lau, L.J., 1986. Functional Forms in Econometric Model Building. In: Z. Griliches and M.D. Intriligator (Editors), Handbook of Econometrics. North-Holland, New York. Leontief, W., 1936. Quantitative input and output relations in the economic system of the United States. Rev. Econ. Statist., 1: 105-25. Leontief, W., 1941. The Structure of the American Economy, 1919- 1929. Harvard Univ. Press, Cambridge. Leontief, W. and Duchin, F., 1986. The Future Impact of Automation on Workers. Oxford Univ. Press, New York. Li, P.-C., 1994. Environmental policy, coal, and the macroeconomy: the impact of air quality regulations on the Pennsylvania economy. Thesis. Dep. Mineral Econ., Pennsylvania State Univ. (unpublished). Li, P.-C. and Rose, A., 1995. Global warming and the Pennsylvania economy. Econ. Systems Res., 7: 151-171. Manne, A.S. and Richels, R., 1992. Buying Greenhouse Insurance: The Economic Costs of Carbon Dioxide Emission Limits. MIT Press, Cambridge, MA. Miller, R.E. and Blair, P.D., 1985. Input-Output Analysis: Foundations and Extensions. Prentice-Hall, Englewood Cliffs, NJ. Montgomery, W.D., 1992. The Carbon Tax, The Environment, and Economic Growth. Paper presented at the Symposium on Enhancing Environmental Quality Through Economic Growth, Washington, DC. Nordhaus, W., 1993. Rolling the DICE: the optimal transition path for controlling greenhouse gases. Resour. Energy Econ., 15: 27-50. OTA (Office of Technology Assessment), 1991. Changing by Degrees: Steps to Reduce Greenhouse Gases. Off. Technol. Assess., Washington, DC. Pennsylvania Coal Association, 1993. Pennsylvania Coal Data 1993. Pennsylvania Coal Assoc., Harrisburg, PA. PRC (People’s Republic of China), Department of Balances of National Economy of the State Statistical Bureau and Office of the National Input-Output Survey, 1990. Input-Output Table of China 1987. China Statist. Publ. House, Beijing. Pyatt, G. and Round, J. (Editors), 1985. Social Accounting Matrices: A Basis for Planning. World Bank, Washington, DC. Robinson, S., Kilkemry, M. and Hanson, K., 1990. The USDA/ERS Computable General Equilibrium Model of the United States. Econ. Res. Serv., Washington, DC.
Global and Planetary Change 11 (1996) 201-221
Rose, A. and Ayres, R., 19%. Economic analysis of greenhouse gas emissions: a smrctural decomposition analysis. Final Report Natl. Sci. Found., Dep. Mineral Econ., Pennsylvania State Univ., University Park, PA. Rose, A. and Stevens, B., 1993. The efficiency and equity of marketable permits for CO, emissions. Resour. Energy Econ., 15: 117-146. Rose, A. and Lin, S., 1995. Regrets or no regrets-That is the question: Is conservation a costless CO, mitigation strategy? Energy J., 16: 67-87. Rose, A. and Miemyk, W., 1989. Input-output analysis: the first fifty years. Econ. Syst. Res., 1: 229-271. Rose, A.Z. , Stevens, El. and Davis, G., 1988. Natural Resource Policy and Income Distribution. Johns Hopkins Univ. Press, Baltimore. Rose, A.Z., Stevens, B. and Davis, G., 1989. Who gains and who
loses from natural resource policy: Distributional information and the public participation process. Resour. Policy, 15: 282291. Rose, A.Z., Stevens, B. and Li, P., 1994. A global marketable permits approach to CO, mitigation: implications for U.S. energy demand. In: E.C. van Ierland (Editor), International Environmental Economics. North-Holland, Amsterdam. Rose, A., Benavides, J., Lim, D. and Frias, 0.. 1996. Global warming policy, energy, and the Chinese economy. Energy Econ., in press. Shoven, J.B. and Whalley, J., 1992. Applying General Equilibrium. Cambridge Univ. Press, New York. U.S. Department of Commerce, 1993. Input-Output Table of the U.S. Economy, 1987. Bur. Econ. Anal., Washington, DC. U.S. Forest Service, 1993. IMPLAN Modeling System. Rocky Mt. Exper. Sm, Ft. Collins, CO.