Contents - Previous - Next

This is the old United Nations University website. Visit the new site at http://unu.edu

8. Macroeconomic costs and other side-effects of reducing CO₂ emissions

Akihiro Amano

1. Introduction

More than 150 nations signed the Framework Convention on Climate Change in June 1992 at Rio de Janeiro, revealing their determination to start coping with the global warming issue. Although the agreements were not as aggressive as had been expected in advance, the extensive guiding principles brought together in Agenda 21 represent a big step forward. One clear message is that more use is to be made of economic measures to combat global warming and other environmental problems. The question we now face is to decide the extent to which we should apply these measures with respect to the abatement of carbon dioxide and other greenhouse gas emissions. The nations that agreed to the Convention have not yet decided about it explicitly except for some environmentally advanced Northern European countries.

In this paper I shall first discuss, in section 2, the magnitudes of the macroeconomic costs of limiting carbon dioxide emissions. By comparing the results of the OECD global model comparison project and those of Japanese studies, I shall point up the importance of understanding the model structures behind these simulation experiments. Besides the macroeconomic costs, economic measures such as carbon taxes (and to a large extent other market-oriented measures as well) will have other side-effects both domestically and internationally. In section 3, some of these problems will also be addressed.

Assessment of mitigation costs alone cannot determine the optimum scale of measures against global warming. Recently, there have been interesting discussions concerning the size of optimal carbon taxes. In section 4, I shall attempt to discern the factors affecting the size of carbon taxes in the optimal abatement path, or the social costs of carbon emissions. Section 5 closes the paper with a summary and conclusions.

2. The macroeconomic costs of reducing CO₂ emissions

When the OECD convened an international workshop in 1991 to compare the simulation results of representative global models on cost estimates of limiting carbon dioxide emissions, one of the objectives was to investigate the reasons such diverse figures had been obtained. Intensive comparative studies of six global models of greenhouse gas emissions culminated in a number of OECD working papers and Economic Studies articles, and many important findings explaining the determinants of macroeconomic costs have been obtained. (See Dean, 1993, for a survey of the workshop.)

Table 8.1 is constructed from the simulation results of four dynamic, long-term models: the Manne-Richels model (MR), Rutherford's Carbon Rights Trade Model (CRTM), the Edmonds-Reilly model (ER), and the OECD GREEN model (GREEN). The figures in the table have been derived from the simulation results involving CO₂ abatement in terms of a 2 per cent reduction in annual rates of increase compared with the business-as-usual scenarios.

The left-hand panel of table 8.1 reports the ratio of percentage reductions in real GDP to those in CO₂ emissions (i.e. percentage GDP losses caused by a one percentage point reduction in CO₂ e missions relative to the baseline), and the right-hand panel reports the ratio of carbon tax rates (measured in dollars per ton carbon) to percentage reductions in CO₂ emissions (i.e. the amount of carbon taxes required to achieve a 1 per cent reduction in CO₂ emissions).

Because these simulation exercises were performed on the same set of exogenous assumptions and on the same method of perturbing the systems, the results turned out to be fairly similar, especially for developed countries. There are a couple of notable differences, however. The Edmonds Reilly model reports relatively larger impacts in the longer term, because backstop technologies do not play a large role compared with other models. The Rutherford model, on the other hand, reports smaller effects, because this model allows for opportunities to trade energy-intensive products and carbon emission rights. Noting these special cases, however, we may say that a 1 per cent reduction in carbon emissions generally requires a 0.02-0.05 per cent decrease in GDP in developed countries. The corresponding carbon tax rates are around US$2-10 per ton carbon (tC).

Table 8.1 The macroeconomic costs of CO₂ emission reduction

Source: Dean and Hoeller (1992).

a. The percentage change in GDP relative to the business-as-usual scenario for a 1 per cent reduction in CO₂ emissions, also relative to the business-as-usual scenario.

b. Carbon faxes required for a 1 per cent reduction in CO₂ emissions relative to the business-as-usual scenario.

Table 8.2 Carbon tax simulations of Japanese models

Source: Amano (1992a,b).

For non-OECD countries the results are somewhat diverse, but I can make two observations. First, all models show relatively larger output effects in these countries, especially in the "rest of the world" region, which includes energy-exporting countries. Secondly, carbon taxes for the former USSR and China are fairly low in the GREEN and Edmonds Reilly models, because these models take into account the subsidized, low domestic energy prices in these countries.

Table 8.2 presents the results of similar simulation studies conducted in Japan. I report this additional information because it clearly shows that the objectives of model-building and the methods of simulation experiments can both influence the results substantially. The figures in table 8.2 are constructed in the same way as those in table 8.1.

We can distinguish two groups in this table. The first group (the Goto and Ban models) obtained comparable results to those of table 8.1 with respect to both GDP/GNP reductions and carbon taxes. The results of the second group, however, are surprisingly similar to each other, but they are much larger than other estimates in either table 8.2 or table 8.1.

Three reasons can explain these differences. First, the models in the second group have been developed by combining econometric forecasting models of the demand-determined type with some form of energy model, and most of them have usually been used for projection and simulation exercises. The role expected of such models is to make precise short- to medium-term projections and not to draw a clear picture of the distant future. On the other hand, computable general equilibrium (CGE) models can generally treat the long-run responses more explicitly and adequately with smoother adjustments in various sectors. As the above results indicate, the long-run and short-run responses captured by these two sets of models can give rise to notable differences.

The second reason for the difference is that short-run models involve temporary deviations from full employment resulting from higher energy prices, which are absent in CGE models by assumption. Of course, this does not mean that short-run adjustment problems are unimportant. There are, however, regular anti-cyclical measures in the policy arsenal, and these measures should be and will be integrated in the policy package at the level of actual implementation.

The third point relates to the treatment of tax revenue. In general equilibrium models, carbon tax revenue is usually recycled to the public to make the carbon tax revenue neutral. However, this is not true for the second group of models in table 8.2; the results reported in the table were based on an assumption of no change in public sector behaviour. This assumption, combined with the demand-determined type of macroeconometric model, can lead to a large decline in output with the imposition of carbon taxes.

These considerations suggest that policy simulation results should be presented and interpreted with care. Analysts try to identify the effects of some factors by isolating the disturbance as far as possible. But the way of isolating an event depends a great deal on how that factor is modelled. Therefore, the results of such simulations must be presented and interpreted with a clear understanding of the model structure.

In fact, actual policies may not be implemented as suggested by the model simulations we have just seen. Assignment of the same percentage reduction in emissions to all countries is not an efficient way of formulating a global or an international policy. An efficient policy would require that the rate of carbon tax be the same for all countries. The way in which tax revenue is recycled may vary from one country to another, reflecting differing public deficits. To say the same thing from a different angle, real policy simulations should be based on combinations of policy measures with realistic policy responses to expected unfavourable side-effects. Ordinary "policy simulations" are often not carried out in this way.

3. Some side-effects of reducing CO₂ emissions

The kind of consideration mentioned at the end of the last section provokes discussion concerning the various side-effects of economic measures to mitigate global warming (such as carbon taxes), including the regressive distributional effects within a country and undesirable effects upon international competitiveness when other countries do not adopt similar measures. There is some empirical evidence to support the first point (see, e.g., Poterba, 1991, and Smith, 1993), but these authors also indicate that there are policy instruments that can offset the socially undesirable distributional consequences of carbon taxes.

The question of international competitiveness seems to involve at least three points. First, when the external costs associated with carbon emissions are internalized, changes in the relative cost structure may lead to alterations in the pattern of the international division of labour. If an international carbon tax scheme is adopted with a uniform tax rate, then an efficient international division of labour will not be distorted, although ordinary short-term adjustment problems will accompany any change in comparative cost structures. Rather, international resource allocation will be improved because the price structure now reflects social costs rather than mere private costs. However, if only a subset of countries participate in this scheme, or if the tax rates vary substantially among countries, then "trade-diversion effects" may result. The supply sources of carbon-intensive goods may shift from more efficient and more benign-to-the-environment countries to less efficient and less benign-to-the-environment countries. In such circumstances, exemption from carbon taxes may be justified. It should be noted, however, that this argument applies only to those industries that suffer from the trade diversion, not to all industries suffering from a loss of international competitiveness.

In this connection, Hoeller and Wallin (1991) show that there exist fairly large differences in implicit carbon tax rates among major OECD countries, and Burniaux et al. (1992) subjoin that the differences are even wider if we consider the world economy as a whole. World energy and carbon uses would be made much more efficient if these distortions could be eliminated.

Secondly, when carbon taxes are introduced in some countries to the extent that world energy prices are depressed, then energy consumption or carbon emissions in other countries not participating in the carbon tax scheme will increase, and this tends to lessen the initial reduction in carbon emissions. This effect, combined with the offsetting influence resulting from trade-diversion effects mentioned above, is called "carbon leakages." Rutherford (1992) reported that the leakage effects are fairly large, especially when carbon limitation becomes very stringent. According to his model simulation, when OECD countries alone attempt to reduce the annual rate of increase in carbon emissions by 3 per cent, the leakage rate, i.e. the proportion of unilateral abatement effects that are offset by the expansion of carbon emissions in non-participating countries, will approach 100 per cent (see Rutherford, 1992). If this conclusion is correct, then any international unilateral action that might affect international energy prices should involve some arrangements to minimize such carbon leakage effects.

On the other hand, simulation analyses performed by the OECD GREEN model suggest that the carbon emission stabilization scheme unilaterally adopted by the OECD countries will lead to carbon leakages of only 2.5 per cent (see Burniaux et al., 1992, and Nicoletti and Oliveira-Martins, 1992). On the average for the period 1990-2050 the greatest reduction in the production of energy-intensive sectors occurs in Japan (- 2.6 per cent) and the smallest reduction in the United States (-0.4 per cent). These results are in sharp contrast to those of Rutherford.

Manne (1993) also examined the extent of carbon leakages with a global model incorporating international trade in crude oil, natural gas, energy-intensive products, and tradable emission permits. According to his simulation results, carbon leakages through the channel of oil trade seem unimportant except perhaps in the initial period. However, international trade in energy-intensive products creates a broad conduit for carbon leakages. The leakage rate starts from around 20 per cent in 2000, increasing to a level slightly above 30 per cent in 2050. Trade in natural gas also raises the leakage ratio in the medium term, but it tends to moderate the leakage in the longer run as world natural gas prices are capped at the backstop level. The overall results seem to fall between Rutherford and the GREEN model results.

Even with these leakage effects, however Manne concludes that unilateral carbon limitations by the OECD nations would be effective in that they could reduce global emissions for some time. At the same time, he also stresses the finding that, beyond 2020 or so, emissions from non-OECD countries will become quite important. These two major conclusions seem to suggest that carbon leakages are at most a medium-term issue, if relevant at all. In the longer teen, many energy-intensive activities will move to the non-OECD region anyway, irrespective of the introduction of carbon taxes in OECD countries, and emission reductions in the non-OECD region will become a central issue. Effective arrangements to contain the vast increase in emissions expected from this region will become imperative under most plausible scenarios.

All three models mentioned above aggregate industries into energy-intensive and energy-non-intensive sectors, but a more disaggregated approach would be needed to verify the results. Quantitative studies of the impacts of carbon taxes upon more disaggregated sectors, such as Jorgenson et al. (1992) for the United States and Kuroda and Shimpo (1992) for Japan, have shown that output responses brought about by the imposition of carbon taxes will be concentrated in a rather small number of energy-related industries such as coal mining, crude oil, electric utilities, gas utilities, and refining. The effects on energy-intensive manufacturing sectors such as iron and steel and paper and pulp products, however, are not as marked as in the energy industries. Therefore, the distinction between energy-intensive and energy-non-intensive sectors does not really indicate the uneven distribution of sectoral impacts upon output. Of course, aggregation does not affect the size of the total impact upon carbon dioxide emissions, so that the implications of simulation results concerning carbon leakages will remain unaffected. However, the distinction masks the differential burden of adjustments among industries which should somehow be taken into account in formulating an appropriate policy package.

Another interesting study applies the GREEN model. Oliveira-Martins et al. (1992) examined the effects of tax exemption of energy-intensive industries, in order to see if such measures can protect these industries from a loss of international competitiveness. Their results show, quite interestingly, that the effects of tax exemption are almost negligible in terms both of leakage rates and of changes in sectoral output. It appears that it is not the loss of competitiveness due to the imposition of the tax but a contraction in the market in question that hits these particular industries.

The third question related to the international competitiveness issue concerns the effects of changes in the terms of trade caused by an international carbon tax scheme. As discussed in relation to the second problem, the international application of carbon taxes, be they unilateral or worldwide, would most probably lead to changes in the international terms of trade of carbon energies in favour of energy-importing countries and against energy-exporting countries, implying large-scale international income transfers. There are not many global models that examine this question, but the OECD GREEN model has shown that the terms of trade effects upon real incomes are of non-negligible magnitudes (see Burniaux et al., 1992). In international negotiations to apply economic measures to mitigate global warming, due consideration will have to be given to this issue.

4. The social costs of CO₂ emissions

One of the important questions remaining unresolved in the discussion of how to cope with the global warming issue is the extent of the desirable, or socially optimal, strategy to reduce greenhouse gases. On the one hand, the group of scientists associated with the Intergovernmental Panel on Climate Change (IPCC) confirmed their earlier recommendation that in order to stabilize climatic change it would be necessary to reduce the current level of emissions of CO₂ by 60 per cent (IPCC, 1992). On the other hand, William Nordhaus has been maintaining in a series of papers that climatic stabilization or even emission stabilization is far from optimal from the socioeconomic viewpoint, and that the socially optimal abatement path is much closer to the uncontrolled path (Nordhaus, 1990a,b, 1992a,b). According to his view, percentage rates of reduction of greenhouse gases along the optimal path will be as low as 10 per cent in around 2000 and 14 per cent in around 2100. The levels of carbon taxes will also be moderate along the optimal path, ranging from about US$6/tC in 2000 to a little above US$20/tC in 2100.

William Cline (1992), on the other hand, considers that a more aggressive policy of stabilizing current CO₂ emissions at an annual rate of 4 billion tons of carbon (GtC) would be justified if (a) future benefits are given higher weights by means of a lower social rate of time discount, and if (b) the risk-averse stances of policy makers are taken into account. Although Cline's conclusions are not based on an optimization model, they are derived from a detailed global cost benefit analysis. If we reconstruct his cost-benefit model, we can calculate the carbon taxes required for the aggressive policy to obtain US$50/tC in 2013, US$100/tC in 2020, US$200/tC in 2025, and US$250/tC in 2054 and after. These rates are much higher than those of Nordhaus.

At an OECD/IEA international conference, Fankhauser and Pearce (1993) reported that their estimates of the social costs of CO₂ emissions, measured as the discounted sum of future incremental damages, are US$20/tC in 1991 -2000, US$23/tC in 2001-2010, US$25/tC in 2011-2020, and US$28/tC in 2021-2030. These estimates fall between those of Nordhaus and Cline, although Fankhauser and Pearce did not present estimates beyond 2030.

In this section I shall examine the factors affecting the magnitude of the shadow prices or social costs of carbon dioxide emissions by means of a small economy-climate model of the Nordhaus type (see the appendix at the end of the chapter for a brief description of the model).

I first performed simulation experiments in order to substantiate the wide differences between Nordhaus's results and those assuming stabilization of CO₂ emissions or of temperature rise. The first five columns of tables 8.3-8.7 present cases where (a) a 3 per cent annual social discount rate is used, as in Nordhaus's analysis, except for Case O' as will be explained below, (b) the IPCC's central estimate of 3°C is used for the climate sensitivity parameter (i.e. the temperature increase at the benchmark condition of doubling carbon dioxide concentration in the atmosphere relative to the pre-industrial level), and (c) the damage function is such that the damage parameter (i.e. the percentage reduction in world GDP at the benchmark climate of 2 x CO₂) is 1 per cent and the function is quadratic.

In these simulations, world output is influenced by climate change through the damage function, which reflects both the macroeconomic costs of emission control and the damage arising from a temperature increase (or the benefits of preventing temperature increase through emission control). The "Uncontrolled" case, however, refers to a situation where these cost-benefit interactions between climate and the economy are completely neglected. In what follows I shall call this the "Business as Usual" (BaU) case.

As expected, the characteristics of Case O are very similar to those of Nordhaus. Optimal percentage reductions in carbon emissions start from 7 per cent in 2000 and remain below 20 per cent all the time. Carbon taxes are also low, starting from US$5/tC and rising to US$13/tC in 21)50 and to US$25/tC in 2100. Reflecting such low emission control. the pattern of temperature rise is very similar to that of the BaU scenario. In other words, a large-scale reduction in CO₂ emissions would not be optimal, and the optimal control path under these conditions almost implies the maintenance of the status quo as far as global warming is concerned.

Table 8.3 Percentage reductions in CO₂ emissions

Table 8.4 Carbon taxes (US$/tC)

Table 8.5 Carbon emissions (GtC)

Table 8.6 Temperature rise (°C)

Table 8.7 Percentage change in world GDP relative to "business-as-usual" scenario

Continue

Contents - Previous - Next