Nature © Macmillan Publishers Ltd 1997

Stephen H. Schneider andLawrence H. Goulder

In thinking about ways to achieve a reduc-tion in the emissions of greenhouse gases, itis crucial to distinguish between the timingof emissions abatement and the timing ofpolicy action. As we discuss below, there arecompelling arguments for implementingonly relatively small CO2 mitigation mea-sures in the short term and delaying signifi-cant abatement to the more distant future,but this does not justify the absence of policyaction now. On the contrary, it is vital to havea short-term abatement policy to bringabout low-cost reductions in CO2 emissions,even when most of those reductions willoccur in the distant future.

We believe that a carbon tax is the mosteconomically efficient and administrativelyflexible instrument for policy action now.Under a range of plausible economicassumptions, we show here that a carbon taxis a better instrument for generating a cost-effective abatement profile than subsidies toresearch and development for alternative,low-carbon sources of energy.

Postponing abatementLast year, Wigley, Richels and Edmonds1

published a widely cited, influential articleon the appropriate timing of emissionsabatement. They examined several alterna-tive models for mitigation pathways culmi-nating in the same long-term concentrationof CO2 and concluded that, in general, over-all abatement costs are kept to a minimum ifthe bulk of CO2 abatement takes place in themore distant future rather than soon.

Wigley et al. offered four explanations fortheir conclusion. The first is the positivereturn on capital: capital growth would meanthat fewer resources need be put aside todayto fund future abatement. The second is thatthe capital stock for energy production anduse may be long-lived. Delayed abatementwould allow expensive production assets tobe gradually replaced with fewer carbon- orenergy-intensive new technologies after the older stock has reached the end of itseconomic life. The third is technologicalprogress: new, energy-efficient technologieswill be discovered and developed, hence the“availability of low-carbon substitutes willprobably improve and their costs reduce overtime”1. And fourth, carbon emitted sooner isexposed to natural removal processes forlonger, so delayed abatement permits a largertotal of cumulative emissions to produce thesame long-term concentration target.

The main idea expressed by Wigley et al.,that society can economize on the costs ofachieving targets for reduced emissions ofgreenhouse gases by deferring most abate-ment to the future, has won many adherents.But others (see, for example, ref. 2) contendthat delay could be costly, as a commitmentto reduce emissions now would be an impor-tant catalyst for technological progress thatwill ultimately lead to lower costs of abate-ment. Much of this disagreement can beovercome by distinguishing between emis-sions abatement and abatement policy.Although there are strong arguments tojustify delaying the bulk of emissions abate-ment, we believe there are compelling arguments for short-term implementationof government policies to set in motion theeconomic adjustments necessary to bringabout low-cost reductions in emissions.

A carbon tax todayClimate-related damage from accumulationof CO2 is a cost that, without governmentintervention, is not incorporated in the mar-ket price of fuels3. A carbon tax, if set appro-priately, would allow market prices to reflectthe full social cost of climate damage and givepurchasers the incentive to economize onfuel use. In addition, because the tax wouldcause prices of carbon-intensive consumergoods to rise relative to other goods, it wouldgive individual consumers the incentive torely more heavily on less-carbon-intensivegoods and services, and producers the incen-tive to develop alternatives.

Of course, other policies could be used todiscourage or restrict the use of carbon-intensive fuels, for example, direct limits or caps, fuel-efficiency standards and man-dated technologies or equipment. Therehave been many economic comparisons ofthese alternatives4,5, but most tend toembrace the carbon tax as the most econom-ically efficient and flexible instrumentbecause it imposes fewer informationrequirements on policy-makers; it providesdynamic incentives; is relatively inexpensiveto administer; and is relatively easy to adjustin response to new information (essential inclimate assessment). Moreover, a carbon taxwould bring governments revenues thatcould be used to finance cuts in ordinaryincome taxes, thereby helping to avoid in-efficiencies associated with disincentives towork or to save. This ability to generate rev-enue would make a carbon tax much moreefficient than limiting carbon use6.

An appropriately scaled carbon taxwould also help to dictate the most efficient

time-profile for abatement. We agree withWigley et al.1 that it is more cost-effective to defer to the future the bulk of carbonemissions abatement (relative to the path of emissions under an unconstrained ‘business-as-usual’ model). But this does notjustify avoiding carbon taxes in the shortterm.

Economic analysis indicates that carbontax rates should be set according to the‘marginal environmental damage’ from CO2

emissions. Most analyses imply a carbon taxrate that rises with time3,7,8 to encourage thelevel of abatement to increase. Even a con-stant carbon-tax rate, though less economi-cally efficient, is consistent with a risingtime-profile for abatement because, as newtechnologies for emissions (or fuel) abate-ment are discovered and implemented, theprofit-maximizing amount of emissions-abatement expands, even if the tax rate isconstant. Thus, there is no contradictionbetween introducing carbon taxes now and acost-effective time-profile for abatement.

Introducing the carbon tax now could bea key factor in inducing the technologicalchange that justifies deferring most abate-ment to the future. Making carbon-basedfuels more expensive provides incentives toresearch alternative energy-supply options.As emphasized in ref. 2, a carbon tax can alsostimulate technological change by promot-ing “learning-by-doing”. In the context ofclimate change, this means that experiencewith processes that reduce CO2 emissionsmay lead to the discovery of cheaper ways toreduce emissions. In our view, introducingthe carbon tax now would usefully exploitthis phenomenon.

Research subsidies or carbon taxes?Does the fact that a carbon tax may stimulateresearch mean there is no justification forintroducing a research and developmentsubsidy as part of climate-change policy? Ageneral economic principle is that govern-ments should apply the policy instrumentmost closely related to a particular ‘marketfailure’. As noted above, the central marketfailure in this case is the climate damage asso-ciated with combustion of carbon-basedfuels. A research subsidy does not directlydeal with this market failure because, unlikea carbon tax, it does not directly alter theprices of carbon-based fuels. If there were noother market failures to be concerned about,a subsidy would be unnecessary: the carbontax alone would reduce the accumulation ofCO2 in a cost-effective manner.

But there is a second market failure —

Achieving low-cost emissions targetsThere has been much discussion about ways to reduce the rate of atmospheric accumulation of carbon dioxide andother greenhouse gases. What are the key economic and policy issues concerning the timing of such efforts?

commentary

NATURE | VOL 389 | 4 SEPTEMBER 1997 13

Nature © Macmillan Publishers Ltd 1997

commentary

14 NATURE | VOL 389 | 4 SEPTEMBER 1997

failure in the market for research anddevelopment. It is well known that privateinvestment in research and developmentcan generate ‘spillover benefits’ that areenjoyed by parties other than the investor.Not all knowledge can be kept as privateknowledge. Under these circumstances,companies tend to underinvest in researchand development, mainly because they donot take into account the full social value(including the spillover benefits) when theymake the investments. A government sub-sidy can correct this market failure, ideallyby lowering a company’s costs enough toallow it to expand its investment in researchand development to the socially optimallevel.

Does this argument imply that a carbontax should be accompanied by a subsidy forresearch into alternative energy supplies? Itdepends. Economic theory suggests that asubsidy is appropriate wherever there aresignificant spillovers. If such spillovers arisein connection with virtually all industrialresearch investment, then the most eco-nomically efficient policy response is abroad-based subsidy for a wide spectrum of industries. On the other hand, if suchspillovers are specific to investment in alter-native energy supplies, then there is a basisfor a more targeted subsidy. Although thereare strong theoretical and empirical reasonsto support a carbon tax, the case for a tar-geted research and development subsidy isless clear.

We have developed an economic simula-tion model for the United States9 which webelieve is the first large-scale, general equi-librium, economic model to take intoaccount incentives to invest in research anddevelopment, knowledge spillovers, and thefunctioning of research and developmentmarkets. The estimated costs of reducingcumulative CO2 emissions by 15 per cent inthe 100 years after 1995 is shown in Table 1. Aresearch subsidy alone never offers thecheapest way to meet the target reduction incumulative emissions and indeed can bemany times more costly than the other poli-cies. Results from our model are sensitive toparameters that are highly uncertain; thus,we emphasize the qualitative pattern fromTable 1, not specific numbers. These simula-tions support our view that a carbon tax isessential for cost-effective reductions of CO2

emissions, and that this tax should beaccompanied by a research subsidy onlywhen there are spillover benefits fromresearch and development.

Two qualifications deserve mention.First, if research and development marketsare already highly inefficient (for example,distorted by previous subsidies or taxes),then gauging the costs of new subsidies thatshould be applied to low-carbon energysources becomes more complicated. Depen-ding on the array of pre-existing subsidies,

the costs of a combination of carbon tax andsubsidy might be higher or lower than sug-gested in Table 1, but this does not affect thegeneral principle of our argument. Second,our model focuses on private-sector research(which may be publicly subsidized). Thus,our simulations do not directly address theeffectiveness of research that is directlyundertaken by the public sector.

Political and other considerationsA carbon tax policy is likely to be less popularthan subsidies for research into energy tech-nologies, even though a policy consistingonly of subsidies would not be the cheapestway to reduce emissions. A subsidy is morefavourable to producers but a carbon tax is more attractive to general taxpayers(although some low-income groups areconcerned about its potential regressivity). A research subsidy needs to be financedthrough other tax revenues, whereas a gov-ernment can ‘recycle’ carbon-tax revenuesto the benefit of general taxpayers (includinglower-income groups). Discussions of policyoptions need to take into account theseeffects on the tax-paying public.

We have concentrated here on the issuesof cost-effectiveness and economic effici-ency. Of course, in the political arena thereare advocates for other policy principles,such as the precautionary principle or theprinciple of stewardship. Our purpose is notto debate the relative merits of alternativecriteria for policy choices, but simply showthat a ‘do-nothing’ policy is difficult to justify, even according to a policy criterionendorsed by most economists.

A further important issue is the existenceof uncertainty. One view is that it is prema-ture to initiate policy action now, given theuncertainties about the extent of climatechange and associated damage, whereasanother is that postponing policy actionwould be much more costly in future ifaction were ultimately required. Recentanalyses indicate that it is worthwhilestarting policy action now, despite theuncertainties, provided that there are not

prohibitively high ‘sunk costs’ (unrecover-able one-time costs) of introducing parti-cular policies10,11. A carbon tax policy isattractive because it involves minimal sunkcosts, suggesting that short-term action iswarranted, and because it is flexible enoughto allow adjustments as new informationabout climate change and the effectivenessof policy becomes available12.

In conclusion, we support the view that itmay be cost-effective to defer most CO2

abatement, but we believe that policy actionis needed now for cost-effective futureabatement. A carbon tax (or other flexible,direct policy confronting the effects of fossil-fuel combustion) is an essential element ofgreenhouse policy. A research and develop-ment subsidy could be a useful complementto a carbon tax when there are research and development market failures. But thecase for a subsidy (in terms of economicefficiency) rests primarily on spilloverbenefits from general research (or otherresearch and development market distor-tions) and not on the prospect of environ-mental damage from atmospheric build-upof CO2.Stephen H. Schneider and Lawrence H. Goulder arerespectively in the Departments of BiologicalSciences and of Economics, Stanford University,Stanford, California 94305-5020, USA. Bothauthors are co-appointed in the Institute forInternational Studies at Stanford.e-mail: shs@leland.stanford.edu

1. Wigley, T. M. L., Richels, R. & Edmonds, J. A. Nature 379,

240–243 (1996).

2. Grubb, M. Energy Policy (February 1997).

3. Nordhaus, W. D. Am. Econom. Rev. 72, 242–246 (1991).

4. Hahn, R. & Stavins, R. Ecol. Law Quart. 18, 1–42 (1991).

5. Office of Technology Assessment Environmental Policy Tools: A

User’s Guide (US Govt Printing Office, Washington DC, 1995).

6. Goulder, L. et al. Rand J. Econom. (in the press).

7. Peck, S. C. & Teisberg, T. J. Energy J. 13, 71–91 (1992).

8. Goulder, L. H. & Mathai, K. Working Paper, Department

of Economics (Stanford Univ., 1996).

9. Goulder, L. H. & Schneider, S. H. Res. Energy Econom.

(submitted).

10.Manne, A. & Richels, R. Buying Greenhouse Insurance

(MIT, Cambridge, Massachusetts, 1996).

11.Pindyk, R. Working Paper, Department of Economics

(MIT, Cambridge, Massachusetts 1996).

12.Schneider, S. H. Envir. Modelling Assessment (in the press).

Table 1 Costs of 15% reduction in CO2 emissions 1995–2095

Model Carbon tax Targeted Carbon tax plus Carbon tax plusalone R&D subsidy targeted R&D subsidy broad R&D subsidy

alone of 10% of 10%

1. No spillover from R&D 0.94 8.52 1.02 1.18

2. Spillovers from0.66 5.98 0.60 0.78R&D only by alternative

energy industry

3. Spillovers from R&D 1.03 9.55 1.09 0.81investment by all industries

Figures are percentage reductions to the present value of GDP. All simulations involve carbon tax rates thatincrease at a rate of 5 per cent annually to the year 2075 and remain constant thereafter. The carbon taxprofile is the lowest path of (rising) tax rates that leads to the 15 per cent reduction in cumulativeemissions relative to the baseline model. Model 1 evaluates costs when there are no spillovers fromresearch and development (R&D) investments. In this case, the cheapest way to attain the 15 per centabatement target is through a carbon tax alone. Model 2 assumes that there are significant spillovers frominvestments in R&D by the ‘alternative energy’ industry (the non-carbon-based fuel industry). In this case,the combination of carbon tax and R&D subsidy to alternative energy is the more cost-effective way toattain the target. Model 3 assumes all investments in R&D involve significant spillovers. In this case, theleast-cost policy involves a combination of carbon tax and broad subsidy to R&D.