addressing the climate change problem). Instead of a threat to jobs, reducing the economy's dependence on fossil fuels can be seen as an investment and jobcreation opportunity, because of the new equipment and technologies that will be required. The conversion can be accomplished without any net loss of jobs; the role of policy is to minimize transition costs and to ensure that any such costs do not fall disproportionately on narrow segments of the population such as coal industry employees.

It is also an exaggeration to claim that fossil fuel emissions reduction policies would cause massive capital flight from the industrialized economies. Annual U.S. capital outflows, including intercompany debt and reinvested earnings, has been averaging about 7% of total annual gross domestic capital formation. The flow of equity capital alone is considerably smaller, averaging about 2% of annual gross domestic capital formation over 1992-1996. (See Bureau of Economic Analysis, Survey of Current Business, various issues. Direct investment components were current cost values adjusted to 1992 dollars using the BEA's GDP deflator. For definitions and discussion of the data, see the article by Mataloni (1995) in the Survey of Current Business.)

A large portion of these capital flows are to countries that would face changes in the relative price of fossil fuel energy similar to those experienced by the United States under a climate protection treaty. On an historical cost basis, the cumulative U.S. direct investment position in Canada, Europe, and Japan was 68% of the total in 1995 (Bach 1997). Given the magnitudes of these numbers, it is not plausible that gradually-phased-in changes in the price of fossil fuels undertaken as part of a climate treaty could cause massive capital flight from the United States.

Economic research has found no convincing evidence that environmental regulations have a significant effect on businesses' locational decisions or competitiveness. The most recent and extensive survey of the literature on the relationship between environmental regulation and the competitiveness of U.S. manufacturing concluded that "there is relatively little evidence to support the hypothesis that

environmental regulations have had a large adverse effect on competitiveness, however that elusive term is defined....[S]tudies attempting to measure the effect of environmental regulation on net exports, overall trade flows, and plant-location decisions have produced estimates that are either small, statistically insignificant, or not robust to tests of model specification" (Jaffe et al. 1995, pp. 157-8). A great many factors influence decisions on where to build new plants, including the availability and quality of the labor force, proximity to output markets and input sources (i.e., transportation costs), the locational preferences of management, tax considerations (other than environmental taxes), and political stability. Pollution regulations are far down on this list. Some work suggests that properly designed environmental standards may even bolster productivity (Porter 1990, 1991; Porter and van der Linde 1995; Goodstein 1997). Thus, even if the Protocol negotiated in Kyoto has differential emissions reduction requirements for developed and developing countries, it will not cause a flight of new investment to the developing countries.

B. "Bottom-up" Cost Estimates

The top-down estimates are premised on the idea that reductions in GHG emissions can only be purchased at the expense of other goods and services. An alternative picture of the economic effects of GHG emissions abatement is given by the bottom-up studies. The IPCC surveyed the literature and found a large body of evidence that substantial emissions reductions could be accomplished at a net gain to the economy (IPCC 1996c). In the studies surveyed by the IPCC, an emissions reduction on the order of 25% from the base year level could be achieved at zero net average cost.

13

Since publication of the IPCC report, several new studies estimating the positive economic potential of energy-efficiency improvements have appeared. The

13. The 25% figure is the median of the different values obtained from studies with an ending point between 2000 and 2030. Some of the studies did not include options all the way up to the intersection point of the x-ax ́s. This means that larger reductions could be achieved for negative or zero net cos:

most comprehensive of these, Scenarios of U.S. Carbon Reductions: Potential Impacts of Energy-Efficient and Low-Carbon Technologies by 2010 and Beyond, was prepared by researchers at five of the national laboratories-Oak Ridge National Laboratory. Lawrence Berkeley National Laboratory, Pacific Northwest National Laboratory, National Renewable Energy Laboratory, and Argonne National Laboratory (Interlaboratory Working Group on Energy-Eficient and Low-Carbon Technologies 1997). This study examines four key sectors (buildings, transportation, industry, and electric utilities) in detail, and concludes that it would be possible to reduce carbon emissions to roughly 1990 levels by 2010 at "net costs to the U.S. economy...near or below zero in this time frame."'"'

A second major new bottom-up study was carried out by a consortium of the Alliance to Save Energy, the American Council for an Energy-Efficient Economy, the Natural Resources Defense Council, the Tellus Institute, and the Union of Concerned Scientists (1997). This report, aided Energy Innovations: A Prosperous Path to a Clean Environment, finds that the United States could follow an "Innovation Path" that by 2010 would lead to "a national energy system that, compared to the Present Path, reduces net costs by $530 per household, reduces global warming CO2 emissions to 10 percent below 1990 levels, and has substanually lower emissions of other harmful air pollutants." Finally, an adaptation of the U.S. Energy Information Agency's National Energy Modeling System (NEMS) model has been developed building in more dynamic assumptions about market transformation and behavioral change than originally contained in the NEMS model (Hoffman and Sylvan 1996). Running the NEMS model with these assumptions results in forecasts of GHG emissions in 2015 reduced by 13% to 39% from the NEMS baseline (the 39% reduction corresponds to a 21% reduction from 1990 emissions levels), with a GDP gain of between 0.3% and 0.5%.

Companies and individuals around the world already are earning profitable returns by investing in energy-saving technologies such as modern fluorescent lighting

14. The 1990 emissions levels could be reached by 2010 with a carbon price of $50 tonne, assuming that this price signal would be accompanied by a vigorous national commitment to develop and deploy cost-effective energyefficient and low-carbon technologies....(along with utility sector investments........" (Interlaboratory Workung Group 1997).

systems, variable-speed motors, computer-controlled HVAC (heating, ventilation, and air-conditioning) systems, new CFC (chlorofluorocarbon)-free refrigeration and cooling systems, and improved building design. Corporate leaders are increasingly recognizing the potential for earning money while making significant GHG reductions, as indicated by participation in voluntary energy-saving programs in the United States (such as the EPA's Green Lights and Energy Star initiatives, or the DOE/EPA Climate Wise program), and by the recent announcement by Keidanren, the multisector Japanese business group, to cut its CO2 emissions 10%20% from 1990 levels by 2010 (Global Environmental Change Report 1997).

C. Towards Resolution of the Question of Costs

The existence of these two strands of literature on the cost of GHG reductions poses a problem of scientific methodology for economists. Both sets of estimates cannot be right, yet there is no clear consensus on how to make the two types of calculations consistent. Therefore, it is useful to step back from the divergent estimates in order to gain some perspective on why, in this instance, economists disagree.

The results obtained by top-down models are strongly dependent on the assumptions built into the models. A recent meta-analysis of the estimates of the costs of climate protection that were obtained from 162 runs of 16 of the most widely used models shows that almost all the variation in predicted economic impacts is accounted for by differences in eight key assumptions. These assumptions include: the availability and cost of a non-carbon "backstop" technology, the efficiency of the economy's response to price changes, the degree of inter-fuel and product substitutability, how many years are available to achieve the specified CO2 reduction target, whether reducing CO2 emissions would avoid some economic costs of climate change, whether reducing fossil fuel combustion would avoid other (non-climate) air pollution damages, how the carbon tax revenues are recycled into the economy, and whether "joint implementation" options are available or not (Repetto and Austin 1997).

15. Joint implementation" refers to a country such as the United States obtaining credit towards its GHG reduction targe: by financing an emissions-reducing project in another country.

The top-down studies also have built-in assumptions about how technology changes that tend to overstate the costs of GHG mitigations. For example, if future technological change, particularly in energy efficiency, is limited to historical rates, model estimates will not capture the effect of policies that could speed up the rate of technical progress. Investments in research and development (R&D) have high payoffs, and it is well-established that that new knowledge is a "public good" that will not be produced in socially optimal quantities without government support. A private company can capture only a portion of the economic gain that results from successful R&D activity, and this is why estimates of the social rate of return to R&D are higher than market rates of return on ordinary private investment. Table 1 (see page 26) gives a summary showing rates of return to R&D estimated by various investigators.

The average social rate of return to R&D from this table is 63.8%; the private rate of return to R&D is 31.8%." Mansfield (1991) states, "[d]ozens of economists working independently with quite different sorts of models and entirely different kinds of data have found that the social rate of return from industrial innovations and R&D has been very high, frequendy 40 percent or more. This is a remarkable fact, and one that policy-makers should recognize." This is why there is such a broad consensus that government should subsidize basic research. Government initiatives to increase the rates of innovation and diffusion of technologies that would help reduce greenhouse-gas emissions could substantially change the values of the technical change parameters embedded in the top-down models, and could have a positive impact on the growth performance of the economy as a whole."

16. Both endpoints of ranges were included in computing the averages.

17. It can be argued that increasing R&D in the energy-efficiency sector could drain scarce scientific and technical resources from other, potentially more productive, R&D activities. (See Goulder and Schneider (1996) for a good development of this argument.) The seriousness of this problem depends on the extent to which (1) the total R&D budget for the economy is fixed, and (2) the allocation between energy-efficiency R&D and other R&D is already optimized. Neither of these conditions would appear to hold now. Indeed, we know that it is possible to increase aggregate A&D substantially; this is a policy decision having mainly to do with the funding of graduate educa.en for scientists and engineers, and with the availability of jobs and equipment for those researchers upon completion of their degrees. The time lag for beginning to see the effects of such an effort would be three years at the mos. We have the historical experience of the post-Sputnik push that demonstrates the feasibility (and benefits) of an increase in society-wide R&D. Nor is the national allocation of R&D effort optimal. Public research collars are not allocated on the basis of their expected rate of return, even excluding the very large expenditures on military R&D (which con. tribute to military preparedness but do not have a direct economic return at all).

46-495-7