Costs of prevention

A very simple representation of preventive costs is used in PAGE, reflecting the current confused state of knowledge, with results from top-down and bottom-up studies often disagreeing by orders of magnitude.

Even if no measures are taken t o combat global warming, emissions of each gas in each region will still change over time because of population change, economic growth, and policy decisions unrelated t o the greenhouse effect. For instance, carbon dioxide emissions in the rest of the world will grow with population growth and industrialization, while CFC emissions in the EC will decline because of commitments t o protect the ozone layer.

This is modeled in PAGE by defining an emission profile for each gas in each region that can be achieved without incurring any costs.

In this study, the zero cost profile is taken t o be the 'no-action' preventive policy. Future economic growth rates and other policy measures are not completely known, so a probability distribution is defined for each gas t o show the effect of this uncertainty on the profile.

Three probability distributions for each gas express the cost of any re- duction below this profile. The first parameter is the annualized unit cost in 1990 ECU per ton of the cheapest control measures, the second gives the range of cutbacks over which these cheap measures can be employed, and the third gives the additional annualized unit costs of any cutbacks larger

Table 5. Costs of carbon dioxide control measures in the EC.

In the other world regions, an additional single probability distribution for each region multiplies the cost parameters for preventive measures in

?No methane control strategies are considered in this study, since t h e technical potential for control is not yet established. C F C control is included, and t h e costs are based broadly upon Nordhaus (1991b), but they d o not amount t o more than a few percent of carbon dioxide control costs in any analysis year.

'This range of values is an attempt t o incorporate the wide divergence of views ex- pressed in t h e literature. Top-down macroeconomic models have tended t o report high control costs, as in Nordhaus (1991b) and Manne and Richels (1991b). Bottom-up studies such as M i l s et al. (1991) have found costs much closer t o the lower end of the range t h a t we use. T h e results of Barker et al. (1993), which incorporate the effects of recycling carbon tax revenues, and seem t o demonstrate very low or even negative preventive costs, came too late t o be included in this study.

Table 6. Discounted global emission control costs of the "aggressive" pre- ventive policy, by time perioda (trillion E C u b ) .

Low Mean High

value value value

"Aggressive" total 0.8 2.7 5.7

of which

1995-2010 0.2 0.6 1.3

2011-2050 0.5 1.6 3.3

2051-2100 0.1 0.5 1.1

aEmission control costs of the "no-action" preventive policy are essentially zero through- out.

b ~ n late 1992, 1 ECU is approximately equal to US$1.

higher costs of construction and civil engineering works in the EC's high wage economy.

Within PAGE, probability distributions are also defined for each pollu- tant in each region other than the EC describing the percentage range of cutbacks over which the cheaper measures can be employed. The rationale for this is that whereas the costs of the various control measures are un- likely t o vary greatly across the world, reflecting only the influence of high local manufacturing costs etc., the range of cutbacks over which cheaper measures can be applied will depend strongly on the particular patterns of industrialization in a region, and so will vary from region t o region and gas t o gas.

T h e emission control costs that result from these inputs are shown in Table 6, aggregated across all world regions and discounted back t o 1990 a t a 5% discount rate. Costs of between 1 and 6 trillion ECU, even when spread over the next 110 years, are not negligible; they are of the same order of magnitude as one year's output from the whole of the EC.'

T h e costs of control are in fact concentrated in the first half of the next century; over half of the discounted control costs come in the period 2011- 2050, and over 20% come before 2010. This contrasts with the discounted global economic impacts, shown previously in Table 4, where only about 6%

of the total came before 2010, and over half occurred after 2050.

The obvious comparison to make is between the costs of control in Table 6 and the reduction in economic impacts that the "aggressive" preventive policy might bring. Table 7 shows this comparison. At a 5% discount rate, the mean results show the extra costs of control, a t 2.7 trillion ECU, just

'Gross Domestic Product in the EC in 1987 was 3.7 trillion ECU (Eurostat, 1989).

Table 7. Reduction in economic impacts and extra costs of control world- wide: "aggressive" preventive policy versus "no action" (trillion ECU a t a 5% discount ratea).

Low Mean High

value value value ^{- - ~ ~}

-Reduction in economic impacts 0.5 2.0 3.4

Extra costs of control 0.8 2.7 5.7

"In late 1992, 1 ECU is approximately equal to US$l.

outweighing the 2 trillion ECU benefits from the reduction in economic im- pacts. Only if the costs of control are towards the lower end of the range of estimates, or if a lower discount rate than 5% per year is thought t o be appropriate, would the direct economic benefits from the "aggressive"

preventive policy outweigh its costs.

It would be possible t o pursue this comparison further; it turns out, for example, that in the EC region and the rest of the OECD, even on the mean results and at a 5% discount rate, the costs of the "aggressive" preventive policy are justified by the reduction in direct economic impacts. However, as we commented earlier on the results in Table 4, it would not seem sensible t o try t o tackle global warming by prevention alone; however draconian the cutbacks in emissions, some warming and consequent economic impacts will occur, and so adaptation to that warming will need t o be considered. We turn now t o the method of handling adaptive policies in PAGE, and the results obtained from including a package of adaptive measures in the study.