affecting the marginal tax rates). There are, of course, numerous ways in which tax revenues can be used. These include reducing budget deficits; reducing marginal rates of income, payroll, corporate, or other taxes; granting tax incentives to preferred activities; or increasing the level of government expenditures. The costs of the tax will vary widely, depending on how the revenues are recycled.

Table 9.4 shows the range of GDP losses associated with a carbon tax rising from $15 per ton [tonne?] in 1990 to $40 per ton [tonne?] in 2010, with alternative methods of recycling the revenues. The analysis is performed using four different economic models of the U.S. economy (two macroeconomic models and two general equilibrium models). The first alternative recycling method, the lump sum tax cuts, is in the neutral manner described above.

Note that the GDP costs of the carbon tax vary considerably depending on how the revenues are recycled. In some models, the costs are more than offset by tax policies that encourage investment. On the other hand, one model suggests that the GDP costs of the tax would be increased over the neutral case if the revenues were used simply to reduce the government budget deficit. Although the alternative of increasing government expenditures was not examined, it is possible that such a policy would increase the GDP costs of the tax over the neutral case (see Nordhaus, 1994).

9.2.1.2 A Review of Bottom-up Studies

Bottom-up research in the U.S. and Canada has tended to suggest, as elsewhere in the world, that significant decreases in CO2 emissions are possible without great cost to the economy. In this sense, their results differ from those of most top-down research, the latter suggesting substantial economic costs to CO2 emission abatement. The methodology discussions in Chapter 8 reviewed some of the key reasons for these differences.

9.2.1.2.1 Variations in bottom-up technology and policy assumptions

Like top-down studies, bottom-up studies result in a wide range of reduction cost estimates. Compared with top-down studies, the structure of formal models in bottom-up analysis is generally less important for the results. Instead input assumptions are dominant. Inspection of available studies shows that for a given time horizon and geographic region, divergent results arise mainly from differences in two factors:

the quality of the technical analysis from which supply curves for energy efficiency and supplies from cogeneration and renewables are derived;

the assumed effectiveness of policy instruments in mobilizing the economically cost-effective resource potential.

Differences in technical analysis. Of the technical factors, the most important ones appear to be the level of detail in analyzing and representing supply curves for technology options, notably those for demand-side efficiency improvements. Bottom-up studies that rely on more detailed and comprehensive assessments of these options will tend to arrive at larger efficiency potentials and lower costs of saved energy than less detailed studies. Even then, uncertain baseline data for equipment use and efficiency in individual end uses leave some room for disagreement among technologists.

The treatment of administrative policy costs can also be important. Most bottomup analyses assume that energy efficiency standards would be the main policy tool. These have negligible administrative costs, but such costs can be more substantial in some types of incentive programmes. State-of-the-art studies give a differentiated treatment.

Depending on the scope of the study, additional no-regrets opportunities may be quantified. These are lower generating costs from utility regulatory reforms, savings in acid rain and other pollution control expenditures from reduced fossil fuel use, the costreducing effects of monetizing environmental externalities (other than climate change), and increased cost-effectiveness of nonfossil options when removing fossil fuel price subsidies.

Most bottom-up studies do not quantify the feedback effect from lower energy demand on fuel prices or the further effect of lower fuel prices and energy service costs on cost-effective energy efficiency levels and energy demand. As a result, bottom-up studies may at once underestimate feasible economic savings from carbon reductions and overestimate the amount of emission reductions that market transformation policies can bring.

Differences in assumed policy effectiveness. All bottom-up scenarios of future energy demand assume that policy intervention can at least partially shift the investment behaviours of consumers and firms from historically suboptimal patterns to economically optimal choices. At one end of the spectrum, it is assumed that policies will shift every replacement purchase or expansion of end-use equipment that will occur over the time period studied. For short time horizons (10-20 years), this assumption implies that a complete shift to efficient equipment will be achieved within one cycle of capital turnover. This assumption is almost certainly too optimistic.

At the other end of the spectrum, some bottom-up studies take a very pessimistic view in which current political difficulties are assumed to limit the more widespread application of market transformation policies. As a result, the savings potential is estimated to be low.

A compromise position is found in studies using longer time horizons (30-40 years or more). Over these long periods, most capital goods will be replaced more than once, and many several times. Here, least-cost efficiency levels can be achieved within the time

horizon, even if policies do not shift all or most investments the first time around. The assumptions about the effectiveness of policies is thus more realistic in these studies.

9.21.22 Key study assumptions and results

In this section we survey some of the key assumptions and results in the studies for the U.S. and Canada. Unfortunately, there has not yet been an effort to undertake a harmonized comparative exercise of U.S. bottom-up models, as has been the case with U.S. top-down models. However, a survey of the approaches at least allows for some tentative observations.

An overview of the main assumptions and results of the major U.S. bottom-up studies is provided in Tables 9.6, 9.7, and 9.8. Most studies test the sensitivity of their results to changes in GDP growth rates. Energy prices are generally assumed exogenously, although some studies have attempted to incorporate the feedback effects of efficiency measures on energy prices. RIGES, for example, assumes low-cost fossil energy in 2030 as a consequence of successful implementation of energy efficiency [REFERENCE? OR PREFERABLY PROVIDE REFERENCES FOR RIGES AND FFES IN THE TABLE.]. In contrast, Carlsmith et al. (1990) assume high rates of near-term energy price increases.

Table 9.8 reconfigures the reported results from Tables 9.6 and 9.7 in a manner that facilitates interpretation for policy making. Results are now reported in terms of the percentage reduction below the level of base year emissions that was found to have zero net cost. The base year variously ranged from 1985 to 1990. The reduction cost is reported as average cost. For studies where negative net average costs were reported, feasible zero-net cost reductions are shown with a "greater than" sign to indicate that the results reported do not include reduction options up to the intersection point with the xaxis.

The following reductions were found to result in zero net costs:

By the turn of the century: 0-21% (median of 11%)

By about 2005-2010: >0-26% (median of 13%)

By about 2015-2020: >23-58% (median of 41%)

By about 2025-2030: >61-82% (median of 72%)

The studies were not normalized in terms of such factors as economic growth, structural change, fuel prices, technology costs, or policy effectiveness. As a result, the above

reduction ranges are wider than a coordinated analysis based on uniform assumptions would have found.

Despite these wide variations, the figures show a consistent pattern: Within the 1990-2030 time frame, progressively larger reductions become feasible at zero net cost as time horizons grow longer. And for any given reduction target, mitigation costs decline as longer adjustment periods are allowed.

This pattern reflects two major factors: first, bottom-up studies identify large energy efficiency potentials that are not exploited in the reference case, due to the high transaction costs. The studies assume that these high transaction costs will be reduced through suitable policy interventions, such as efficiency standards, least-cost utility planning, and other information and financial incentive programmes.

Second, most energy efficiency improvements are introduced at the "economic optimal" rate of capital stock turnover, with the more long-lived capital goods lasting twenty to thirty years or more. Due to this synchronicity, cost-effective efficiency improvements "keep coming."continue to occur. As a result, bottom-up studies estimate "transition benefits" rather than "transaction costs" for such technology shifts.

Third, cheap cogeneration opportunities are assumed in a number of studies to make an important further negative-cost contribution. Compared to separate generation of electricity and heat, this supply-side technology is found to have significant costeffective resource potentials. They are typically not included in the reference case, since business-as-usual regulatory regimes do not sufficiently control monopolistic [monopsonistic? It's a case of buyer domination.] utility buy-back practices. Like most energy efficiency improvements, cogeneration technologies are already commercially available and thus add to the near- to medium-term "no-regrets Ypotential.

When cost estimates are compared for a given reduction target, the results of the bottom-up studies listed in Tables 9.6, 9.7, and 9.8 lie within a reasonably close range, despite considerable differences in assumptions. Most cost estimates for a 20% reduction in U.S. CO2 emissions by the year 2010, for example, range between -0.6 and +0.5% of GDP (see Table 9.7). As noted in section 9.2.1.2.1, this difference is explained at least in part by the nature and level of detail of the input assumptions used in the specific bottom-up analysis. The RIGES study, for instance, assumes 100% penetration of markets by all technologies shown to be cost-effective in an engineering/economic analysis. In contrast, Carlsmith et al. constrained their model to link all market penetration of efficient technologies to consumer behaviour parameters (expressed as price elasticities) and policy constraints (expressed as efficiency standards). This latter approach did not lead to 100% market penetration by efficient technologies.

Bottom-up studies for Canada were compiled and reviewed by Robinson et al. (1993) in the study, Canadian Options for Greenhouse Gas Emission Reduction

(Robinson, et al, 1993). Estimates of cost-effective CO2 emission reduction potential by 2010, relative to a reference (or baseline) scenario, ranged from 20% to 40%, with a median of about 23%. Relative to 1988 or 1990, many studies showed savings in energy use or emissions of between 10% and 30%, with a median of about 16%.

9.2.2 Studies of the Costs of Reducing CO2Emissions in Other OECD Countries

Much of the early work on the costs of CO2 emission reduction was U.S.-based and, as a result, tended to be U.S.-focused. More recently, there has been a flurry of analytical activity elsewhere in the OECD, mainly in Western Europe. In general, these countryspecific studies have had relatively shorter time horizons than in the U.S., focusing on the costs of stabilizing emissions in 2000 or a 20% reduction by 2005 (the "Toronto target").

9.2.2.1 A Review of Top-down Studies

Top-down studies of non-U.S. OECD countries have been of two types: those focusing on an entire region or a subset of countries and those focusing on individual nations. In this section, we review the results of both. As with the U.S. studies, the types of policies tested have been limited, for the most part, to taxes on CO2 and energy under alternative domestic fiscal recycling schemes.

9.2.2.1.1 Regional Studies

One notable attempt at a systematic model-comparison of non-U.S. OECD models was conducted by the OECD in the early 1990s (Dean and Hoeller, 1992). The exercise was patterned after the parallel study being undertaken by the Energy Modeling Forum. The two studies used many of the same models and shared a common set of input assumptions. For purposes of the model comparison, the OECD was divided into two regions: the U.S. and "other OECD countries."

The OECD analysis encompassed both the transition and backstop phases. This long-term perspective is useful when examining issues related to the timing of the transition away from fossil fuels and the potential role of technical innovation in lowering overall costs of reducing emissions.

The OECD study examined a range of emission reduction scenarios. Among the more interesting is one in which emissions are permanently held to 1990 levels. Figure 9.7 compares annual GDP losses for three models: ERM, GREEN, and MR [Is MR the Manne and Richels Global 2100 model?]. The models are fairly consistent in their projections of losses in 2010 - between 0.3% and 0.5% of GDP. The convergence is due in part to the standardization of key input parameters. But it is also important to note the aggregation effect. Combining all non-U.S. OECD countries into a single region