Costing non-linearities, surprises, and irreversible events

Department ofBiologicalSciences,Stanford

University,Stanford, CA

94305-5020, USA

Stephen H Schneider, Kristin Kuntz-Duriseti, Christian Azar

Costing non-linearities, surprises, andirreversible events

Pacific and Asian Journal of Energy 10(1): 81-106

Abstract The primary purpose of this paper is to highlight issues thatare crucial when costing climatic impacts, particularly whenthe possibility is allowed for non-linearities, surprises, andirreversible events. The assumptions made when carrying outsuch exercises largely explain why different authors obtaindifferent policy conclusions. Uncertainties become more sig-nificant when projections of climatic impacts are considered.There is uncertainty about how the biosphere will respond tohuman-induced climate change. However, it is clear that life,biogeochemical cycles, and climate are linked components ofa highly interactive system. Non-linearities and the likelihoodof rapid, unanticipated events (surprises) require that costingmethods use a wide range of estimates for key parameters orstructural formulations, and that, when possible, results becast in probabilistic terms rather than central tendencies sincethe latter mask the policy-relevant wide range of potential re-sults such a diversity of approaches implies. Costs need also tobe presented in more numeraires than just monetary ones.This paper recommends that key for authors of scientific as-sessments is transparency of assumptions and the use of aswide a range of eventualities (and their attendant probabili-ties) as possible to help decision makers become aware of thearguments for flexibility of policy options.

S H Schneider, K Kuntz-Duriseti, C Azar82

Valuing climate change underuncertainty

The combination of increasing population andper capita energy consumption will contrib-ute to increasing CO2 (carbon dioxide) andsulphate emissions over the 21 sl century, but

projections of the extent of their increase areuncertain. According to IPCC (Inter Govern-mental Panel on Climatic Change) (1996a),CO2 concentration will double preindustriallevels by the middle of the 21 Sl century, which

is projected to lead to a warming of 1 DC tomore than 5 DC by the end of the 21 sl century.

Warming of 1 DC could have significant impli-cations for species adaptation, whereas warm-ing of 5 DC or more could have catastrophiceffects on natural and human ecosystems, in-cluding serious coastal flooding. The overallannual cost of these impacts in smarket sec-tors' of the economy could run into tens ofbillions of dollars (Smith and Tirpak 1990,IPCC 1996b). Although fossil fuel use con-tributes substantially to such impacts, associ-ated costs are rarely included in the price ofconventional fuels; they are externalized. In-ternalizing these environmental externalities(Nordhaus 1992, IPCC 1996c, Goulder andKennedy 1997) is a principal goal of interna-tional climate policy analyses.

Uncertainties become more significantwhen projectio~s of climatic impacts are con-sidered. The extent of the humap imprints onthe environment is unprecedented: human-in-duced climate change is projected to occur ata rapid rate, natural habitat is fragmented foragriculture, settlements, and other develop-ment activities, 'exotic' species are importedacross natural biogeographic barriers, and theenvironment is assaulted by chemical agents(Root and Schneider 1993). It is, therefore,essential to understand not only how much cli-mate change is likely, but also how to charac-terize and analyze the value of the ecosystemservices that might be disrupted. There is un-certainty about how the biosphere will respondto human-induced climate change. However,

Pacific and Asian Journal

it is clear that life, biogeochemical cycles, andclimate are linked components of a highly in-teractive system.

The primary purpose of this paper is to high-light issues that are crucial when costing cli-matic impacts, particularly when the possibilityis allowed for non-linearities, surprises, andirreversible events. The assumptions madewhen carrying out such exercises largely ex-plain why different authors obtain differentpolicy conclusions. The overall cost of climatechange involves the costs of mitigation, adap-tation, and the remaining damages. Uncer-tainty and the possibility of surprises surroundeach of these components and have a profoundeffect on them. In this paper, first, we discussthe conditions for non-linear events and sur-prises, followed by their importance for thecosting of climate damages. Finally, we con-sider various response strategies, includingadaptation and mitigation.

Imaginable conditions forsurprise

JRate of forcing

The most comprehensive models of a compli-cated coupled system like an ESM (earth sys-tem model) are likely to have unanticipatedresults when forced to change rapidly by exter-nal disturbances like CO2 and aerosols. Indeed,some of the transient coupled atmosphere-ocean models run out for hundreds of yearsexhibit dramatic change to the basic climatestate-radical change in global ocean currents(Mana be and Stouffer 1993, Haywood,Stouffer, Wetherald, et al. 1997, Rahmstorf1999). Stocker and Schmittner (1997) arguethat rapid alterations to oceanic currents couldbe induced by faster forcing rates.

Thompson and Schneider (1982) used sim-plified transient models to investigate whetherthe time evolving patterns of climate changemight depend on the rate at which CO2 con-centrations increase. For slowly increasing CO~build-up scenarios, the model predicts the

of Energy 10(1): 81-106

83Costing non-linearities, surprises, and irreversible events

standard outcome: the increase in tempera-ture at the poles is more than that in the trop-ics. Any change in equator-pole temperaturedifferences creates altered regional climates,since temperature differences influence large-scale atmospheric wind and ocean current pat-terns. However, for rapid increases in CO2concentrations, Thompson and Schneiderfound a reversal of the equator-pole tempera-ture differenc~'in the Southern Hemisphereover many decades during and after the rapidbuild-up of CO2, This would imply unexpectedclimatic conditions during the century or sothe climate adjusts toward its new equilibriumstate. In other words, the faster and harder wepush on nature, the greater the chances for sur-prise-some of which are likely to be damaging.

Clearly, rapid transients or non-linear eventsare likely to cause alterations to higher statis-tical moments of the climate (e.g. week-'to-week variability, seasonal amplitudes,day-to-night temperature differences, etc.).Such rapid or unexpected events are likely tocontradict the 'invariance of higher moments'.Thus, resultant environmental or societal im-pacts are likely to be different from those thatwould occur with smoother, slower changes.The long-term impact of climate change maynot be predictable solely from a single steadystate outcome, but may well depend on thecharacteristics of the transient path; the out-come may be path-dependent. Any exercise,which neglects surprises or assumes transitiv-ity of the earth system (i.e., a path-independ-ent response) is, therefore questionable, andshould carry a warning to users of the funda-mental assumptions implicit in the techniquedependent on steady state results.

Assessment and reporting ofuncertainties

Moss and Schneider (2000) and Moss (thisvolume) note that the term uncertainty canrange in implication from a lack of absolutesureness to such vagueness as to preclude any-thing more than informed guesses or specula-

Pacific and Asian Journal (

tion. Uncertainty results from lack of infor-mation, or is caused by disagreement aboutwhat is known or even knowable. Some cat-egories of uncertainty are amenable to quanti-fication, while others cannot be sensiblyexpressed in terms of probabilities. Uncer~tainty is not unique to the domain of climatechange research. However, in climate research,problems are compounded by additional char-acteristics. These include their global scale,long time lags between forcing and response,low frequency variability with characteristictimes greater than the length of most instru-mental records, and the impossibility of be-fore-the-fact experimental-controls. Moreover,because climate chan~e and other complex,socio-technical policy..ssues are not just sci-entific topics but also matters of public de-bate, it is important to recognize that even gooddata and thoughtful analysis may be insuffi-cient to dispel some aspects of uncertainty as-sociated with the different standards ofevidence and degrees of risk avers-ion/accept-ance that individuals participating in this de-bate may hold. ..

SurprisesA surprise is an unanticipated outcome. How-ever, the IPCC (Intergovernmental Panel onClimate Change) SAR (Second AssessmentReport), defines surprises as rapid, non-linearresponses of the climate system to anthropo-genic forcing, and cites analogies topaleoclimatic abrupt events to demonstrate theplausibility of such a possibility. The SAR alsogives specific examples of such non-linear be-haviours that the authors could envision asplausible (e.g. reorganization of thermohalinecirculation, rapid deglaciation, fast changes tothe carbon cycle).

It would be better to define these as imagi-nable abrupt events. The Working Group I SARconcludes its Summary for Policymakers withthe statement that non-linear systems whenrapidly forced are particularly subject to un-expected behaviour (IPCC 1996a). Of course,

:>f Energy 10(1): 81-106

84 S H Schneider, K Kuntz-Duriseti, C Azal'

the system would be less rapidly forced if deci-sion makers chose, as a matter of policy, toslow down the rate at which human activitiesmodify the atmosphere. To deal with such ques-tions the policy community needs to under-stand both the potential for surprises and howdifficult it is for lAMs (integrated assessmentmodels) to credibly evaluate the probabilitiesof currently imaginable surprises let alonethose not currently envisioned (Schneider,Turner, and Morehouse Garriga 1998).

Valuation of costs of climatedamage

Costing of extreme eventclimate damages

Subjective probability assessments of poten-tial climate change impacts provide a crudemetric for assigning dollar values to certainaspects of ecosystem services. Costs associ-ated with global change can be anticipatedand a preliminary value placed on some of theecosystem services that could be affected.Evaluation of the losses from extreme climaticevents, such as floods, droughts, and hurricanesis one way to assess the costs of climatechange (Alexander, Schneider, and Lagerquist1997).

Cautious projections indicate that a warmerclimate and human activities such as urbani-zation, deforestation, depletion of aquifers,contamination of ground water, and poor irri-gation practices will increase both the fre-quency and intensity of catastrophic floods anddroughts (IPCC 1996a). Humanity remainsvulnerable to extreme weather events. For ex-ample, between 1965 and 1985 floods in theUnited States claimed 1767 lives and causedproperty damage of more than S 1.7 billiondollars. Alexander, Schneider, and Lagerquist(1997) base their estimates on federal expen-ditures because information of private insur-ance losses and costs are unavailable.Ultimately, the effects of these flooc.5 are feltacross a wide range of economic sectors, as

can be seen with the overall cost evaluation ofthe Midwest flood of 1993 (Table 1).

In the 1993 Midwest flood, 9 states and 525counties declared disasters. The estimated fed-eral response and recovery costs included $4.2billion in direct federal expenditures, $ 1.3 bil-lion in payments from federal insurance pro-grammes, and more than $621 million infederal loans to individuals, businesses, andcommunities. In the upper Mississippi Valleystates of Minnesota, Nebraska, North Dakota,and South Dakota, as well as Wisconsin andnorthern Iowa, losses were primarily agricul-tural. In Illinois, central Iowa, and Missouri,major losses occurred in agriculture as a resultof bottomland flooding, but urban areas alsosustained damages. Numerous,impacts of theflooding are still largely unknown, includingcumulative effects of releases of hazardousmaterial such as pesticides, herbicides, andother toxics; effects on groundwater hydrol-ogy and groundwater quality; distribution ofcontaminated river sediments; and alterationof forest canopy and sub-canopy structure. Inaddition, the loss of tax revenue ha$ not beenquantified for the Midwest flood. While not allcosts of the flood can be directly calculated inmonetary terms, both quantifiable and non-quantifiable costs were significant in magni-tude and importance. This event, though notdirectly caused by anthropogenic climatechange, allows a rough estimate of the magni-tude of costs should such climate change causeincreases in extreme weather events. Moreo-ver, similar events in less developed parts ofthe world (e.g. flooding from Hurricane Mitchin Central America) may have caused less ab-solute monetary damages but greater losses interms of human life, infrastructure and thesocial fabric of whole communities, not tomention the much higher percentage loss toGDP. Clearly, it is important to be explicitabout the units of cost (numeraire) being con-sidered in each specific case.

Like floods, severe droughts of the 20th cen-tury have affected both the biophysical and



Table 1 Summary of federal expenditures for the Midwest flood of 1993 (million US dollars)

Missouri Iowa Illinois OtherstatesB TotalMinnesota

141.6291.5152.151.9

128.719.34.5

15.01.0

73.57.65.1

891.8

376.2

189.8

107.7

48.5

9.7

22.8

11.1

15.0

1.2

22.1

4.6

2.1

810.8

446.262.929.8

7.90.34.00.85.00.77.32.26.0

573.1

63.3

197.594.98.4

70.37.41.4

10.00.4

33.35.3

11.8504.0

512.2

290.9

75.1

23.8

12.0

15.2

2.2

19.6

0.7

36.9

12.4

8.3

1,009.3

1,699.91,098.0

500.0201.3253.1

75.0100.064.64.0

146.734.041.2

4,217.8

USDAb

FE MAc

HUDd

Commerce

USACEe

HHS

Education

Labour

National Community

DOT'

EPAS

DOlh

Total

a Denotes combined costs, including those for the states of Kansas, Nebraska, North Dakota, South Dakota, and

Wisconsin; b United States Department of Agriculture; C Federal Emergency Management Agency; d Housing and UrbanDevelopment; e United States Army Corps of Engineers; HHS, Department of Health and Human Services; 'Department

ofTransportaiton; g Environmental Protection Agency; h Department of the Interior.

Source Interagency Floodplain Management Review Committee report to the Administrative Floodplain Management

Review Committee report to the Administrative Floodplain Management Task ~orce, 1994.

perform a more integrated analysis, such asthe cost assessment of future sea level rise alongthe US coasts associated with possible ice capmelting or with ocean warming and the result-ing thermal expansion of the waters. In a prob-ability distribution of future sea level rise by2100, changes range from slightly negativevalues to a metre or more rise, with the midpoint of the distribution being approximatelyhalf a metre (Titus and Narayanan 1996).Many studies have assessed the potential eco-nomic costs of sea level rise along the devel-oped coastline of the United States. For a50-cm rise in sea level by 2100, estimates ofpotential costs range from $20.4 billion (Yohe,Neumann, Marshall, et al. 1996) to $138 bil-lion (Yohe 1989) in lost property-dependingon the levels of adaptation assumed. The fol-lowing sections explore how the costs of pre-vention compare to the losses potentially

socio-economic systems of many regions. Es-timated damage from the 1988 drought in theMidwestern United States shows a reductionin agricultural output by approximately one-third, as well as billions of dollars in propertydamage.

Hurricanes can also cause devastation intens of billions of dollars. Warmer surface wa-ters in the oceans currently produce strongerhurricanes. Other meteorological factors areinvolved, though, that may act to increase ordecrease the intensity of hurricanes with cli-mate change.

Damage assessment is one way to relate thecost of inland and coastal flooding, droughts,and hurricanes to the value of preventing thedisruption of climate stability. In the 1993Midwest flood, for example, Alexander, Sch-neider, and Lagerquist (1997) delineate thecosts of a single event. It is also possible to



sustained by increasing floods and droughtsor by future sea level rises, by placing a valueon climate changes and abatement.

Methods of valuationAssessing the cost of climate requires estimatesof both monetary and non-monetary impacts.While the former is difficult, the latter is evenmore complicated and controversial. The needfor alternative methods of evaluation of theseclimate-related ecosystem services is clearwhen examining preliminary public opinionresponses of global warming. In a controver-sial method called contingent valuation(Goulder and Kennedy 1997), respondents aresurveyed to determine how much they wouldbe willing to pay to prevent a given global cli-mate change scenario or how much money theywould require to permit a given amount ofchange. The difficulties with this type ofvalu-ing of environmental goods and processes areimmense, especially since much of the evalua-tion is subjective. Public opinion depends, inpart, on people's exposure and the level of edu-cation and information about these issues.

In a Southern California study, the contin-gent valuation technique was applied to deter-mine the influence of potential changes intemperature and precipitation resulting fromglobal warming on respondents' willingness topay (Berk and Schulman 1995). Respondentswere provided with a baseline microclimate forthe region before future climate scenarios wereevaluated. For example, for residents living incoastal communities, the baseline climate overthe past ten years was described as having:(1) an average high temperature in summersof75 of, (2) daily high temperature range from70-80 of, with some days having over 90 of,and (3) an average of twenty inches per year ofrain. With these and other scenarios, predictedprobabilities were determined from the re-spondents' willingness to pay for the abate-ment of different mean high temperatures. Inthese scenarios, respondents were willing topay an average of S 140 to offset a mean high

temperature of 1 00 of, while a mean high tem-perature of80 of was worth approximately$100. This represents a 40% increase in will-ingness to pay for a 20 of rise in temperatureand other scenario characteristics. The resi-dents, however, reached a plateau in their will-ingness to pay at about 100 of. They were notwilling to pay much more to prevent 120 ofmean high temperatures than to prevent 110of mean high temperatures (Figure 1). This isdissimilar to the respondents in the Nordhaus(1994a) survey of experts who all assigned ac-celerating damage costs to climate change sce-narios as the change became larger-aplausible assumption given that damages areoften non-linearly larger the further changesare from current means.

However, the actual damages to the Los An-geles basin residents from mean high tempera-tures of 110°F or more would be considerablymore costly than those from 100 of (e.g., given~hat landscaping costs alone are tens of thou-sands of dollars, $140 is a vast understate-ment). Berk and Schulman (1995) cautionagainst taking the dollar valu'es from the

Predicted probability of WTP1.0,0.9-0.8-0.7-0.6-0.5-0.4-

0.3-,0.2L0.10.0 , r , . .

50 60 70 80 90 100 110 120 130Mean temperature provided in scenario (degrees F)

Figure 1 Predicted probabilities of Los Angelessurvey respondents' willingness to pay for theabatement of different mean daily high

temperaturesFrom Alexander et al. 1997. Source: Berk and

Shulman 1995.



survey literally or using them in cost-benefitanalyses, as they confound several sources ofvalue including stewardship and altruism. Inaddition, some of the climate changes are wellabove the range of current scientific estimatesof greenhouse warming (IPCC 1996a). Thesurvey was not done in conjunction with at-mospheric scientists and climatologists whocould provide more realistic climate scenariosor ecologists, public health officials, or otherswho could help the respondents realize whatsuch warming might mean for trees, birds, orpeople. This highlights the difficulty in find-ing acceptable methods to place values on theclimatic components of ecosystem services.Therefore, contingent valuation points out thatpeople are willing to pay to preserve ecosys-tem services, but require additional informa-tion to value, more realistically, climate andother environmental services.

Maler, et al. 1996, Nordhaus 1997, Markandya1999).

The descriptive approach focuses on ob-served market interest rates to ensure effi-ciency. For instance, if the market interest rateis 10% and there is a choice between a cost ofUS $100 today and US $109 next year, thelatter is chosen since US $100 set aside todaywould generate US $110 next year and 'earn'US $1 net. Thus, basing the discount rate onthe observed market interest rate can be seenas a way to guarantee that investments are madein the most profitable projects.

However, the longer the time horizon, theless likely is compensation along the linessketched above (i.e. long-term real rates ofgrowth have been closer to 1 % per year, not10%). This means that one would explicitlyhave to discuss trade-~ffs between consump-tion today and in the distant future, which isthe focus of the second approach.

The prescriptive approach emphasizes thatnormative questions are involved in valuingthe future. Proponents of this method oftenbase the discount rate on the SRTP (social rateof time preference), which includes two mainreasons for discounting-the expectation thatwe are going to get richer in the future in com-bination with decreasing marginal utility ofconsumption (i.e. we get less 'satisfaction' peradditional unit of consumption), and impa-tience, which is often referred to as pure rateof time preference. In the prescriptive ap-proach, the choice of discount rate entails achoice on how the future should be valued.This means that there is no objectively correctway to value the future, rather it is a questionof value judgements.

When applying the prescriptive approach,a normative basis is sought for the pure rateof time preference. Proponents of inter-generational equity argue that the pure rate oftime preference, but not necessarily the dis-count rate, should be zero, since it is difficultto find an argument why individual myopiashould translate into giving lower weight

DiscountingDiscounting plays a crucial role in the eco-nomics of climate change. Changes in this pa-rameter largely explain why authors such asNordhaus (1994b) and Manne, Mendelsohn,and Richels (1995), find optimal emissionsincreasing by a factor of three or so over thenext century whereas Cline (1992), Azar andSterner (1996), Hasselmann, Hasselmann,Giering, et al. (1997), and Schultz and Kasting(1997) find that substantial reductions can bejustified within the framework of cost-benefitanalysis using damage functions similar tothose ofN ordhaus. The reason is that discountrates will eventually reduce future damage coststo negligible present values. Consider a climateimpact that would cost one billion US dollars200 years from now. A discount rate of 5% peryear would make the present value of that fu-ture cost equal to

US $58 000, while a discount rate of 10%per year, would make the present value equalto US $5. There are two approaches to findingan appropriate discount rate: the descriptiveand prescriptive approaches (Arrow, Cline,


88 S H Schneider, K Kuntz-Duriseti, C Azar

where c is per capita consumption, utility func-tions are of constant-relative-risk-aversiontype, and the pure rate of time preference isequal to zero. If future consumption rates fallbelow present levels, e.g. as a result of a cli-mate catastrophe, the future should be valuedhigher than the present.

Despite their limitations, these alternate dis-count methods demonstrate the importanceof the structure of the discount function. Allcost assessments need to explicitly explore theirsensitivity to alternative discounting values andstructures.

Need for probabilitydistributions

Attempts to achieve more consistency in as-sessing and reporting on uncertainties are be-ginning to receive increasing attention. Someresearchers express concern that it is difficultto even know how to assign a distribution ofprobabilities for outcomes or processes lacedwith different types of uncertainties. However,the scientific complexity of the climate changeissue and the need for information that is use-ful for policy formulation requires researchersand policy makers to work together towardsimproved communication of uncertainties. Theresearch community must also remember thatusers ofIPCC reports often assume what theythink the authors believed to be the distribu-tion of probabilities it is not specified. Mossand Schneider (2000) argue (Moss, this vol-ume) that it is more rational for experts to pro-vide their best estimates of probabilitydistributions and possible outliers than to havenovice users make their own determinations.

In developing an estimate of a probabilitydistribution, the first step is to documentranges and distributions in the literature, in-cluding sources of informa:t1on on the keycauses of uncertainty, describing how theranges and distributions are constructed, andclearly specifying what they signify. This shouldinclude attention not only to the central ten-dency, but also to the end points of the range

to the welfare of future generations. Similarviews have been expressed by Sidgwick (1890),Ramsey (1928), Harrod (1948),Rawls (1972),Spash and d'Arge (1989), Broome (1992),Cline (1992), Solow (1992), Price (1993),Eriksson (1994), Khanna and Chapman(1996),Azar and Sterner (1996), Rabl (1996),and Schultz and Kasting (1997).

There is empirical evidence to suggest thatindividuals exhibit 'hyperbolic discounting',i.e. higher (than market) discount rates are usedin the short term and lower discount rates areapplied over the long term (Ainslie 1991). Thisbehaviour is consistent with a common find-ing that 'human response to a change in astimulus is inversely proportional to the pre-existing stimulus' (Heal 1997, p. 339). Azarand Sterner (1996) assume that per capita in-come grows logistically over the next century,and since the discount rate is proportional togrowth rates, declining discount rates are ob-tained.

When valuing catastrophic impacts, thevalue for the discount rate depends on the mag-nitude of the damage. In Ramsey-type opti-mal growth models, the discount rate on goods(in this case equal to the social rate of timepreference) is given by

r = yg + p

where r is the discount rate, y is the relativegrowth rate in per capita consumption, y is thenegative of the elasticity of marginal utility ofconsumption (y> 0), and p is the pure rate oftime preference. In lAMs, generally, r is posi-tive. However, if climate change is really se-vere, such that future income falls rather thangrows, then the discount rate becomes nega-tive, provided r is sufficiently low (Azar andJohansson 1996). The discount factor can beobtained by integrating (1) over time

V(t)=expl - =~u'(co)

.!:..9.--

c(t)'rdt,l (2:J

0

Pacific and Asian Journal of Ener~y 10(1): 81-106


representation of the full range of uncertaintyassociated with the estimate. This has impor-tant implications regarding the extent to whichthe report accurately conveys uncertainties.Moss and Schneider (1999) acknowledge thatsome authors are likely to feel uncomfortablewith the full range of uncertainty, because thelikelihood of a 'surprise' or events at the tailsof the distribution may be remote or essen-tially impossible to gauge, and the range im-plied could be large. Thus, there maybe a casefor providing a truncated range in addition tooutliers for a specific case, provided that it isclearly explained what the provided range in-cludes and/or excludes. If a truncated range isprovided, the likelihood that the answer couldlie outside the truncated distribution, shouldbe specified along with the basis for specifyingsuch possibilities.

Consider the example of using probabilitydistributions to evaluate climate damages. Sev-eral studies suggest that climate change willhave only minor economic impacts, and thatan optima\ policy would, therefore, incorpo-rate only modest controls on greenhouse gasemissions (Kolstad 1993, Nordhaus 1992,Peck and Teisberg 1992). However, many ofthese 'modest controls' conclusions are basedon point estimate values-results that are de-rived from a series of best guesses. This pointestimate method fails to account for the widerange of plausible values for many parameters.Policy-making in the business, health, and se-curity sectors is often based on hedging againstlow probability but high consequence out-comes. Thus, any climate policy analysis thatrepresents best guess point values or limited(truncated) ranges of outcomes restricts theability of policy makers to make strategichedges against such risky outlier events.

Nordhaus (1992) has been criticized for con-sidering only a single damage function and notaccounting for abrupt climate 'surprise' sce-narios. In response to such concerns, N ordhaus(1994a) conducted a survey of conventionaleconomists, environmental economists,

of outcomes, possible outliers, the likelihoodthat outcomes beyond the end points of therange might occur, and the type of distribu-tion of potential outcomes (normal, bimodal,etc.).

The next step might be to quantitatively orqualitatively characterize the distribution ofvalues that a parameter, variable, or outcomemay take. The kind of range and confidenceinterval being constructed, or the types ofpos-sible outcomes included in the range shouldbe clear. For example, do the end points (oroutliers beyond them) include potential knownor imaginable non-linear rapid events? Doesthe 'true' value fall into the specified range witha certain probability? Is the range defined tobe one that includes two-thirds of modelledoutcomes available in the literature?

Finally, an assessment of the central ten-dency of the distribution (if appropriate)should be provided. In developing a best esti-mate, authors need to guard against aggrega-tion of results (spatial, temporal, or acrossscenarios) if it hides important regional'or in-ter-temporal differences. Automatically differ-ent distributions should not be combined intoone summary distribution.

Climate sensitivity is an example (Figure2). Here scientists 2 and 4 offer a different es-timate of range outliers (i.e. values below the5th percentile or above the 95th percentile)for imaginable abrupt events. But the meansand variance of scientists 2 and 4 are quitesimilar to most of the remaining scientists inthis decision analytic survey, except scientist5. This is an example where it would likely beinappropriate to aggregate all respondents'distributions into a single composite estimateof uncertainty since scientist 5 has a radicallydifferent mean and variance estimate than theother 15 scientists.

Truncating the probability distribution nar-rows the range of outcomes described and ex-cludes outliers t~at may include 'surprises'. Atruncated estimate of the full range of out-comes, does not convey to potential users a



atmospheric scientists, and ecologists. Sincethese defy simple quantitative treatment, hetook an alternative approach. Nordhaus useddecision analytic techniques to sample theopinions of a wide range of experts who havelooked at climatic impacts. He asked them toprovide their subjective probabilities as to whatthey thought the costs to the world economywould be from several climate-warming sce-narios. Their median estimates of the loss ofGWP (gross world product) resulting from a3 °C warming by 2090 varies between a loss of0% and 21% of GWP with a mean of 1.9%(Nordhaus 1994a). Even a 2% loss ofGWP in1995, however, represents annual climate dam-age of hundreds of billions of dollars. For a6 °C warming scenario, a median loss of 0.8%-62% with a mean of5.5% was predicted.

This is an example of how estimates ofprob-ability distributions can inform. Although thenumbers themselves are revealing, what is re-ally interesting is the cultural divide acrossnatural and social scientists in his study. Themost striking difference in the study is that thesocial scientists (conventional economists pre-dominantly), believe that even extreme climatechange (i.e. 6 °C warming by 2090) would notimpose severe economic losses. Although thisscenario is usually considered to be a low prob-ability event (Figure 2), it is equivalent to themagnitude of change from an ice age to an in-ter-glacial epoch in a hundred years, insteadof thousands of years. Although with a widerange of uncertainty, most conventional econo-mists surveyed still think climate change eventhis radical would, on average, have only a sev-eral per cent impact on the world economy in2100. In their opinion, most natural services(Daily 1997) associated with current climateare either not likely to be significantly alteredor could be substituted for with only modestharm to the economy.

On the other hand, natural scientists esti-mate the economic impact of extreme climatechange twenty to thirty times higher than con-ventional economists (N ordhaus 1994a;


Roughgarden and Schneider 1999). This groupthinks the damages to the economy (includingnon-market components) from the severe cli-mate change scenario would range from no lessthan several per cent lost up to 1 OO%-the lat-ter respondent assigned a 10% chance of thevirtual destruction of civilization! The 50thpercentile damage estimate from this group isan order of magnitude higher than that of theeconomists. Nordhaus suggests that the oneswho know the most about the economy areless concerned while Schneider (1997a) sug-gests that the ones who know the most aboutthe environment are more worried. The natu-ral scientists, in essence, are less sanguine thathuman ingenuity could substitute for ecologi-cal services. Also, as Roughgarden and Sch-neider (1999) show, there is a positivecorrelation between the absolute amount ofdamage each respondent estimates and thepercentage of total damages each assigns out-side of standard national accounts (i.e. thenatural scientists have higher percentages oftheir losses assigned to the non-market sec-tors). Regardless, either judgment inv.t:>lvesboth economic and ecological assessments, notsingle-disciplinary expertise. Clearly, the evo-lution of interdisciplinary communities cog-nizant of both economic and ecologicalknowledge and belief systems will be needed tomake these subjective opinions more credible andto produce cost estimates that span a reasonablerange oicurrently imaginable outcomes.

Note, however, that despite the magnitudein difference of damage estimates betweeneconomists and ecologists, the shape of thedamage estimate curve was similar-the re-spondents indicated accelerating costs withhigher climate changes. This stands in markedcontrast to the flat willingness-to-pay curve inthe contingent valuation example. The expertsurvey respondents, in general, are at leastaware of non-linearities in climate change dam-ages, unlike the lay public respondents.

The differences in various respondents'estimates of climate damages are cast into


"E ~'" 0

~ -0 .~'" ~ '"<I> '" >E -<I>

"'-0

t:QJ

~ -10QJ -5 0 5 10 15 20

1.

m3

~5 8t-6 I ~ I

7 I ~ I

8 I ~ I

9 I~ I

1.0 I E!3 I

1.1. I E!::3 I

1.2 I~ I

1.3 I E!3 I

1.4 I~ I

1. 5 I E!:3 I

1.6 I Ee3 I -.- -.-J I I I I I J I

-1.0 -5 0 5 1.0 1.5 20

Temperature response given 2 x [COJ (K)

Figure 2 Box plots of elicited probability distributions of climate sensitivity, the change in globallyaveraged surface temperature for a doubling of CO2 (2x[CO2J forcing)Note Horizontal line denotes range from minimum (1 %) to maximum (99%) assessed possible values;vertical tick marks indicate locations of lower (5) and upper (95) percentiles; box indicates intervalspanned by 50% confidence interval; solid dot is the mean and open dot is the medianSource Morgan and Keith (1995)

2.3.86

r~~.3!!J2.72.3

~2Ol~~0.30.2

2.72.0

3.1 1.5

2.9 1.4

2.9 1.8

2.6 .98

3.0 1.4

2.8 1.1

1.9 1.0

3.1 1.0

2.2 1.328 1_0

I~ II I ~ I I

with state change I ~ II ~ I

I~ I

with "surprise"

subjective probability distributions byRoughgarden and Schneider (1999) and thenare used to recalculate the optimal carbon taxrate, using the Dynamic Integrated model ofClimate and the Economy, DICE model, (Fig-ure 3). The natural scientists' damage estimatesprocessed by DICE model produce optimalcarbon taxes several times higher than eitherthe original Nordhaus estimate or those of hissurveyed economists. Clearly, the use ofprobabilistic information, even if subjective,

provides a much more representative picture ofthe broad views of the experts as well as a fairerrepresentation of costs which, in turn, allow bet-ter potential policy insights from this lAM.

Several comparisons between the optimalcarbon tax distributions from Roughgardenand Schneider (1999) and the original DICEmodel can be made, using the data summa-rized in Table 2 and Figure 3.

Comparison of the mode (the most frequentvalue) of the RS (Roughgarden and Schneider)


92 S H Schneider, K Kuntz-Duriseti, C Azi

distribution with the results of the originalDICE model, indicates that DICE is a goodrepresentative of the expert opinion expressedin Nordhaus's survey. The modes of the opti-mal carbon tax distributions are slightly abovezero, close to DICE's recommendation for arelatively light carbon tax. However, the otherproperties of the RS distributions justify dif-

ferent policies. The median and mean of theuptimal carbon tax distributions range fromthree to eight times as high as those featuredin the original DICE run.

The differences between the modes of theRS distributions and their medians and means

f(x)0.014

0.012

0.01

0.008

0.006

0.004

Table 2 Comparison of Monte Carlo simulationresults with the standard DICE model

DICE '92/

~($tonne C)2105

- -21.7366.98109.73517.09

v\0.002

'\ ~I I I I

50 100 150 200Carbon tax ($/ton C) in 2055

Note Surprise values are 95th percentile results

Source Roughgarden and Schneider (1999)oL0

f(x)

0.01410.012 i

f(x)0.035

0.03

0.025

0.02

A 0.01,DICE '92

~0.008 DICE '92

/

~0.0060.0151""

0.01 ~I 0.004 v"'"0.0020.005

0 0 I I , I"' . . . . I I I

0 20 40 60 80 100 0 50 100 150 200 250Carbon tax ($/ton C) in 1995 Carbon tax ($/ton C) in 2105

Figure 3 Probability distributions (f (x)) of optimal carbon taxes in the years 1995, 2055, and 2105 fromMonte Carlo simulationsNote Points showing the optimal carbon taxes calculated by the DICE model are shown for comparisonSource Roughgarden and Schneider (1999)



carbon taxes of DICE and the means of theiroptimal carbon tax distributions are accountedfor by the relatively high survey damage esti-mates, and the remaining two-thirds of the dif-ference can be attributed to the non-linearitiesin the model.

In a sense, the original DICE carbon taxmay be regarded as a point estimate betweenthe mode and median of the distribution ofexpert opinion. However, output from a singlemodel run does not display all the informationavailable nor does it offer sufficient informa-tion to provide the insights needed for well-informed policy decisions. One cannot simplylook at a recommendation for a 'five dollarsper tonne carbon tax' and claim that highercarbon taxes are 'necessarily less economicallyefficient'. In particular, strategic hedging poli-cies to deal with the 95th percentile, high dam-age outcome may well be chosen by policymakers, just as individuals or firms purchaseinsurance against low probability catastrophicoutcomes. Regardless of how risk-prone orrisk-averse is the individual decision maker,the characterization and range of uncertain-ties of the information provided by decisionanalysis tools must be made explicit and trans-parent to policy makers (Moss and Schneider1997). This range of uncer~ainty should alsoinclude estimates for the subjective probabil-ity of varying climatic effects (Morgan andKeith 1995, Nordhaus 1994a), damage esti-mates, discount rates (Cline 1992, Chapman,Suri, and Hall 1995, Azar and Sterner 1996),carbon cycle effects on CO 2 uptake (Kaufmann1997, Schultz and Kasting 1997), and the sen-sitivity of the economy to structural changessuch as induced technological change (Grubb,Ha-Duong, and Chapuis 1974, Repetto andAustin 1997, Goulder and Schneider 1999,Azar and Dowlatabadi 1999). The end resultof any set of integrated assessment modellingexercises will be, as always, the subjectivechoice of a decision maker (Schneider 1997b),but a more comprehensive analysis with un-certainties in all major components explicitly

can be attributed to the preponderance of right-skewness of the opinions given in Nordhaus'ssurvey. Most respondents, economists andnatural scientists alike, offer subjective prob-ability distributions that were "right skewed".That is, most of the respondents consider theprobability of severe climate damage ("nastysurprises") to be higher than the probabilityof"pleasant surprises". These long, heavy tails(which Roughgarden and Schneider label 'sur-prise' in Table 2) pull the medians and meansof the distributions away from the modes. Wetake the 95th percentile results from the RSdistributions as representative of these tails.The 'surprise' estimates for optimal carbontaxes in Table 2 are at least twenty times thelevel of those projected by DICE for the threeyears calculated (1995,2055, and 2105).

Two different effects cause these differences.First, the means of these distributions (4.04%and 11.22% of the GWP damage for 3 DCwarming and 6 DC warming, respectively) aremuch higher than the damag~ estimates usedin DICE (1.33% and 5.32%). Thus, the simu-lation study of Roughgarden and Schneideruses more pessimistic damage functions thanthat of the original DICE model. Second, thenon-linearities of the model will, on average,push optimal carbon taxes even higher. Intui-tively, damage functions derived from thesedamage distributions will never give far moreoptimistic results than those with the originalDICE damage function, but they will occa-sionally result in far more pessimistic out-comes. These occasional 'catastrophic' damagefunctions will lead to a relatively pessimisticexpected value of output. In other words, thesignificant chance of a 'surprise' causes a muchhigher level of'optimal' abatement, relative tothe original DICE formulation.

In addition, Roughgarden and Schneideranalyse the effects of the relative severity ofthe average survey damage estimate versusthose of the non-linearities of the DICE modelin a probabilistic analysis. Approximately one-third of the difference between the optimal



One reason why monetary evaluation be-comes controversial is that the results dependon the level of income of those affected. Forinstance, the VOSL (value of a statistical live)is, according to conventional valuation meth-ods, based on the willingness to pay for in-creased safety. Fankhauser (1995) estimatedthe VOSL at US $1 500 000 and US $100 000in developed and poor developing countries,respectively. Thus, the social cost of the deathof 15 000 people in Honduras becomes equalto the social cost of the death of 1000 peoplein the Netherlands. This example shows thatthe measure of social cost does not appropri-ately reflect the seriousness of an impact.

Alternatively, as a consequence of the stand-ard assumption about declining marginal util-ity with respect to income, a loss of US $1 to apoor person is worth more than that to a richperson. This may provide a rational for intro-ducing weight factors to give higher weight tocosts that affect poor countries. Climate cost-ing studies where this approach has been takeninclude Ayres and Walters (1991), Azar andSterner (1996), Fankhauser, Tol, and Pearce(1997),Azar (1999), and Johansson-Stenman(1999). It may be shown that the additionalweight given to losses in poor countries maycompensate for the lower value attached tolosses of human lives, so that equal weightedVOSL across countries is obtained.

However, let us put these difficulties asideand assume that these valuation methods areuncontroversial. Assume also, just as an ex-ample, that climate change would cause dam-age in Bangladesh equal to 80% of its GDP(gross domestic product), or roughly 0.1 % ofglobal GDP. If the global economy were grow-ing at 2% per year, this assumed impact onBangladesh would show up as merely a delayin global income by less than 3 weeks. In po-litical terms, however, such an event would beconsidered more severe than a three-week de-lay in global income growth. This leads us tothe question whether we have the right to tradecosts of emission reduction in countries (e.g.

categorized and displayed will hopefully leadto a better-informed choice.

It is clear from the Nordhaus studies thatknowledgeable experts from a variety of fieldsadmit to a wide range of plausible outcomes-including both mild benefits and catastrophiclosses - in the area of global environmentalchange. This condition is prone to misinter-pretation by those unfamiliar with the widerange of probabilities most scientists attach toaspects of global climate change. In an inter-disciplinary enterprise like the costing of cli-matic impacts or mitigation policies to be usedas inputs to integrated assessment of globalclimate change problems, it is necessary toconsider a wide range of possible outcomes,along with a representative sample of the sub-jective probabilities that assessment groups likethe IPCC believe accompany each of thosepossible outcomes.

Which of the scientists, natural or social,Nordhaus interviewed are closer to the truthmay one day be empirically determinable, butfor the next decade or so, at least, the differ-ences will remain paradigmatic. However, onepolicy-relevant certainty is that the optimalcarbon tax calculated using damage estimatesfrom the surveyed natural scientists is dramati-cally larger than the tax calculated using dam-age estimates from the surveyed socialscientists (Roughgarden and Schneider 1999).

Distributional impactsHow should we value catastrophic impacts,such as the destruction of entire economieswith tens of thousands losing their lives, theorphanage of children, and the collapse of civilinstitutions and normal civil life? Under suchcircumstances, conventional valuation meth-ods become increasingly difficult to use andthey even break down at some point. The im-pact of Hurricane Mitch on Honduras andNicaragua provides one example of how cli-mate-related events may have severe conse-quences on local economies and how difficultit is to value such impacts in monetary terms.


Costing non-linearities, surprises, and irreversible events 95

son's opinion depends on his/her ability to payfor it. This may (or may not) be acceptable forday-to-day transactions over goods and serv-ices, but it is at best politically contentiouswhen the transactions involve the livelihoodand fates the multitudes. Furthermore, thismethod, which could "be labelled 'one dollarone vote', clashes with the principles of de-mocracy (one person, one vote), especiallywhen crucial, path forming societal decisionsare being taken.

more efficient end-use energy technologies)with large-scale losses of lives and humanhealth. Thus, in most conventional cost-ben-efit analyses, the concept of social costs is blindto distributional issues, which is one of the coreissues in the climate change debate(Munasinghe 1999). This points to the neces-sity of using several numeraires, only one ofwhich may be monetary terms, when present-ing the costs of climate change.

Finally, it should also be noted that distri-butional issues are important not only whenassessing impacts. The distribution of incomeis also a major determinant of the efficient so-lution. It is often said that distributional is-sues can be separated from questions ofefficiency. The task for economists has thenbeen to find the efficient solution, and then letpolicy makers take care of distributional im-pacts. However, the assumption that equity andefficiency can be separated only holds underthe assumption that small (marginal) distri-butional changes are considered. Consider theconstruction of a huge dam, which is expectedto yield social benefits of a billion US dollarsat the expense of forced migration of one mil-lion peasants. Assume also that the WTPagainst the project of the peasants would onlybe US $100 per capita. Thus, the project wouldbe 'more efficient'. However, if the farmerswere richer, then they might have been able toexpress a WTP at US $1500 per capita, whichwould have made the dam construction economi-cally inefficient considering 'full social costs'.

This illustrates how the distribution of in-come affects what is considered efficient. It issimply because the farmers are poor, that itbecomes efficient to carry out a project thatmight have severe negative impacts on theirlivelihood. This stands in contrast with thepolicy position against dam construction onthe grounds that it is t4e poor and vulnerablewho can be expected to suffer for its construc-tion (e.g. the projects along the N armada riverin India or the Yangtze river in China). Theexample shows that the importance of a per-

Five numeraires: monetaryloss, loss of life, quality of life(including coercion to migrate,conflict over resources,cultural diversity, loss ofcultural heritage sites, etc.),species or biodiversity loss,and distribution/equity

Any comprehensive attempt to evaluate thesocietal value of climate change should includesuch things as loss of species diversity, loss ofcoastline from rising sea level, environmentaldisplacement of persons, change in incomedistributions, and agricultural losses. The en-vironment also possesses intrinsic worth with-out a specified market value, such as itsaesthetic appeal, which suggests that the envi-ronment should be treated as an independentvariable in utility. This is what is meant by 'ex-istence value'-a priority is placed on preserv-ing the environment, even if it is not intendedto be personally experienced. This is in addi-tion to the 'option value' of the enviTonmeni:,which we may want to preserve for our possi-ble personal use in the future. There is littleagreement on how to place a dollar value onthe non-market impacts of climate change,such as the loss of human life, biodiversity, orecosystem services.

Addressing this, Nordhaus (1994a) askedhis expert panel to separate their subjectiveprobability estimates of climate damages intomarket (standard national accounts) and non-



market damages, such as the value of lost spe-cies, value of lost wetlands from sea level rise,or the costs from conflicts that might be in-duced by the creation of"environmental refu-gees" (Myers 1993) or any of the othernon-market amenities. Economists and natu-ral scientists assigned different fractions ofdamages to the non-market sector, which par-allels the difference in degree of concern overclimate change. Roughgarden and Schneider(1999) find that most respondents who hadestimated large damages placed the bulk ofthem in the non-market category (natural sci-entists), and those with low estimates had as-signed low damages to non-market values(economists). This raises a major issue aboutthe dimensions of damages, which need evenfiner subdivision than the market and non-market binary characterization.

It is essential for analysis of costs of climatechange impacts or mitigation strategies to con-sider explicitly alternative numeraires and tobe as clear as possible which is being used andwhich omitted. Moreover, before any aggre-gation is attempted (e.g. cost-benefit optimi-zation strategies), authors should firstdisaggregate costs and benefits into severalnumeraires and then provide a 'traceable ac-count' (Moss and Schneider 2000) of how theywere re-aggregated. Such transparency is es-sential given the normative nature of valua-tion of various consequences characterized bythe five numeraires.

Rapid changes andadaptability

Natural variability maskstrends and delays adaptation

The assumptions underlying climate changescenarios determine to a large degree, the im-pacts that specific climatic change scenariosare predicted to have on agriculture, coastlinesor forestry. For example, some analysts(Mendelsohn, Nordhaus, and Shaw 1996;Mendelsohn, Morrison, Schlesinger et al.

2000) employ "hedonic methods" by usingcross sectional measure as a proxy to estimateadaptation responses to climate change overtime. However, such static analytic methodsneglect transient dynamics, irreversibilities andhigher moments like changes in seasonality andvariability (see Schneider 1997b for a criticismof the use of "ergodic economics" to modelclimate change over time). One of the majordifferences in estimates of climatic impactsacross different studies is how the lAM treatsthe adaptation of the sector under study. Forexample, it has often been assumed that agri-culture is the most vulnerable economic mar-ket sector to climate change. For decadesagronomists have calculated potential changesin crop yields from various climate change sce-narios, suggesting some regions now too hotwould sustain heavy losses from warmingwhereas others, now too cold, could gain(Rosenzweig, Parry, and Fischer 1994, Smithand Tirpak 1990). But N Rosenberg(Rosenberg and Scott 1994) has long arguedthat such agricultural impact studies imp,lic-itly invoke the' dumb farmer assumption'. Thatis, they neglect the fact that farmers do adaptto changing market, technology, and climaticconditions. Agricultural economists like JohnReilly (Reilly, Baethgen, Chege et al. 1996)argue that such adaptations will dramaticallyreduce the climate impact costs to market sec-tors like farming, transportation, coastal pro-tection, energy use. Ecologists and socialscientists, however, often dispute this opti-mism, since it neglects such real world prob-lems as people's resistance to trying unfamiliarpractices, problems with new technologies,unexpected pest outbreaks (Ehrlich, Ehrlich,and Daily 1995), or the high degree of naturalvariability of weather. The latter is likely tomask the slowly evolving human-induced cli-matic signal and discourage farmers from risk-ing anticipatory adaptation strategies based onclimate model projections.

Clairvoyant adaptation is challenged bythe noisy nature of the climatic system. It is

Pac:ific: anri Ac;ian JolJrnal of Fnprpv 10(1)' R1-106

97Costing non-linearities. surprises, and irreversible events

countries at least). Rather than millions ofdisaggregated decision makers at the farmlevel, therefore, there may be three or four or-ders of magnitude smaller numbers of deci-sion makers. In essence, the problem inmodelling adaptation rests on how to incor-porate human behaviour via a set of decisionrules into the structure of models so as to makethem more 'actor-oriented'. Decision makerswho turn to such later generations of lAMs tohelp inform them about the costs of climatechange must be aware of the controversial na-ture of assumptions about adaptation behav-iour of various actors, which often lurk invisiblyin different impact assessment studies.

The case of coastal flooding is a good ex-ample of how incorporating climatic variabil-ity can significantly reduce the damagereduction potential adaptive activl~mightotherwise have offered if high levels of naturalvariability did not plague climate changetrends. West and Dowlatabadi (2000) deviseda set of decision rules by which coastal dwell-ers would chose to rebuild, remain in place, orabandon c04stal structures, based on the ran-dom occurrence of storm surges superimposedon a slowly rising sea level trend. The 'noise' ofsuch random storm surge events substantiallyalters the adaptability behaviour of coastaldwellers relative to those clairvoyant agentswhose decision rules do not include the mask-ing effects of climatic variability.

doubtful that those in agriculture or situatedalong the coast will invest heavily in order toadapt their practices so as to pre-empt before-the-fact climate model projections, rather thanreact to actual events. One can only speculateon whether or not agricultural support insti-tutions, the research establishment particu-larly, will be influenced by such projections.The high natural variability of climate is likelyto mask any slowly evolving anthropogenicallyinduced trends-real or forecasted. Therefore,adaptations to slowly evolving trends embed-

i ded in a noisy background of inherent vari-ability are likely to be delayed by decadesbehind the slowly evolving global change trends(Kaiser, Riha, Wilks, et al. 1993, Schneider

I 1996, Morgan and Dowlatabadi 1996,I Kolstad, Kelly, and Mitchell 1999). Moreo-I ver, were agents to mistake background vari-

ability for trend or vice versa, the possibilityarises of adaptation following the wrong set ofclimatic cues, and setting up a major systemmalfunction. In particular, agents might bemore influenced by regional anomalies of therecent past in projecting future trends. Theymay be unaware of the likelihood that recentanomalous experience in one region may wellbe largely uncorrelated with slowly buildinglong-term trends at a global scale or may bepart of a transient response that will reverselater on. In addition, unwarranted compla-cency may result from the inability to foreseenon-linear events.

It is doubtful that millions of disaggregateddecision makers (farmers in this example) willrespond uniformly or quickly to forecasts ofglobal climatic changes from lAMs. On theother hand, one of the technological adapta-tions that could mitigate climatic impacts onagriculture is seed development to cope withaltered climates. But, there are only a smallnumber of seed companies capable of alteringthe genetic character of crops and marketingthese better-adapted strains on a large scale tofarmers (in OECD [Organisation of EconomicCo-operation and Development] -like

Passive versus anticipatoryadaptation

Schneider and Thompson (1985), in anintercomparison of climate change, ozone de-pletion, and acid rain problems, differentiatepassive adaptation (e.g. buying mor,e waterrights to offset impacts of a drying climate)from anticipatory adaptation. They suggest, asa hedging strategy, investing in a vigorous re-search and development programme for low-carbon energy systems in anticipation of thepossibility of needing to reduce CO2 emissionsin the decades ahead. The idea is that it would



be cheaper to switch to systems that were bet-ter developed as a result of such anticipatoryinvestments made in advance. Such active (an-ticipatory) forms of adaptation (e.g. buildinga dam a few metres higher in anticipation of analtered future climate) have been prominentin most subsequent formal assessments of an-thropogenic climate change (National Acad-emy of Sciences 1991). Nearly all modernintegrated assessments explicitly (Rosenberg1993, Rosenzweig, Parry, and Fischer 1994,Reilly, Baethgen, Chege, et al. 1996), or im-plicitly (Mendelsohn, Nordhaus, and Shaw1996, Mendelson, Morrison, Schlesinger, etal. 2000) attempt to incorporate (mostly pas-sive) adaptation. While these studies shouldbe applauded for attempting to recognize andquantitatively evaluate the implications ofadaptive responses on the impact costs of cli-mate change scenarios, serious problems withdata, theory, and method remain. A wide rangeof assumptions should be part of any attemptedquantification of adaptation (Carter, Parry,Harasawa, et al. 1994). Moreover, as repeat-edly argued, both costs and benefits of climatechange scenarios treated by any integrated as-sessment activity should be presented in theform of statistical distributions based on a widerange of subjective probability estimates ofeach step in the assessment process (Yohe1991, Morgan and Dowlatabadi 1996, or Sch-neider 1997b).

Although we do not know which target iswarranted, one popular strategy is to createflexibility - in timing and instruments - todayso that as many options as possible remainopen. Toth (see this volume) presents a rangeof alternative decision frameworks. These is-sues were central in the debate about timing ofemission abatement.

Mitigation strategies and(optimal) carbon taxes

Decision makers face in strategic choices theclimate change debate. Ifwe do not slow downinitially, it may be more costly or more difficult,to slow down fast enough if the risk of a cata-strophic event materializes. Suppose, for in-stance, only minimal CO2-abatement policiesare put in place over the next decade or two. Ifit suddenly turns out that climate change is orcan be expected to be much more severe thaninitially estimated, it will be more costly, or per-haps even impossible to meet certain targets.

The timing of emissionreductions

To meet the ultimate goal of the United Na-tions Framework Convention on ClimateChange, Wigley, Richels, and Edmonds (1996)claimed that it is cost-efficient to defer emis-sion reductions a couple of decades, since thiswould give time to develop carbon-free tech-nologies and to avoid a premature phase-outof the existing capital stock. Their position waschallenged on several grounds (Austin 1997,Grubb 1997, Ha-Duong, Grubb, andHourcade 1997, Schneider and Goulder 1997,Azar 1998, Azar and Dowlatabadi 1999). Forinstance, Yohe and Wallace (1996) and l;Ia-Duong, Grubb, and Hourcade (1997) usedstochastic optimization techniques to deter-mine the optimal hedging strategy under short-term uncertainty about which climate targetto meet. Uncertainty was assumed to be re-solved by 2020. Ha-Duong, Grubb, andHourcade (1997) suggested that short-termabatement is justifiable on economic grounds,whereas Yohe and Wallace were less convinced.The main reason for their diverging results isthat Ha-Duong, Grubb and Hourcade consid-ered a symmetric probability distributionaround 550 ppm, with a 2.5% probability thateven a 400 ppm target had to be met, whereasYohe and Wallace chose an uncertainty rangeas high as 550-800 ppm.

Thus, one ma.y conclude that once we wantto keep low stabilization targets within reach,it is 'optimal' to have substantial short-termemissions. One main driving factor for this re-sult is that it will be costly to opt for higheratmospheric stabilization targets and then



change the energy system at a fast rate if se-vere climatic trends unveil themselves. Unfor-tunately, the full effect of this is not seen in thestudy by Ha-Duong, Grubb, and Hourcade(1997) since they assume that the 400 ppmtarget may not even temporarily be overshot.This means that the model has to start abatingin order to avoid that from happening, regard-less of the probability that this target eventu-ally has to be met.

dollars per tonne (1995 rate). The gross costs(i.e. the costs before accounting for environ-ment-related benefits of abated CO2) ofa speci-fied carbon tax are higher with ITC thanwithout ITC because of the explicit inclusionof the opportunity costs ofR&D. However, thiscomparison assumes no prior subsidies forR&D in any industry, no knowledge spillovers,and that all prior inefficiencies in R&D mar-kets are absent. In general, these efficiency oroptimality assumptions are not met in theeconomy. R&D market failures can be cor-rected by public sector investments like R&Dsubsidies to correct the market failure (Sch-neiderand Goulder 1997). Finally, if there wereserious prior inefficiencies in R&D marketssuch that the marginal benefit ofR&D is muchhigher in alternative energy sectors than inconventional, carbon-based sectors or therewere 'no regrets' energy system inefficiencies,then ITC can imply lower gross costs thanwould occur in its absence. There is a criticalneed to formally acknowledge the wide rangeof plausible cost estimates that arise when pa-rameter value uncertainties are combined withstructural assumptions and normative choicesimplicit in various numeraires.

Perspective on the costs ofmeeting the climate target

There is a widespread concern that CO2 con-trol will impose catastrophic economic costs.Nordhaus (1990) warns 'that a vague premo-nition of some potential disaster is, however,insufficient grounds to plunge the world intodepression'. Nordhaus and other top-downmodellers often find the costs of meeting strin-gent CO 2 control targets in trillions of dollars.Manne and Richels (1997), for example, esti-mate the global present value costs (using a 5%per year discount rate) of meeting a 450-ppmtarget to be as high as 4-14 trillion U~ dollars.Other top-down modellers report similar costestimates. In absolute terms, this is a consider-able cost and thus may create the impressionthat we cannot afford to reduce CO" emissions.

The 'cost' of a carbon taxSchneider and Goulder (1997) have developedan economic simulation model for the US,which takes into account incentives to investin research and development, knowledgespillovers, and the functioning of R&D mar-kets, to estimate the costs of reducing cumula-tive CO2 emissions by 15% in the 100 yearsafter 1995. By allowing energy R&D to com-pete with other economic sectors in a highlyaggregated general equilibrium model of theUS economy, Goulder and Schneider (1999)postulate that a noticeable carbon tax is likelyto dramatically redistribute energy R&D in-vestments from conventional to non-conven-tional sectors, thereby producing ITCs(induced technological changes) that lowerlong-term abatement costs. They also demon-strate that there may be an opportunity costfrom ITC. The key variable in determining theopportunity cost is the fundability of humanresources. If all knowledge generating labouris fully employed, then increased R&D in non-carbon technologies will necessarily come at acost to reduced labour in conventional activi-ties. Unfortunately, most integrated assess-ment models to date do not include anyendogenous ITC formulation (or if they do, itis included in an ad hoc manner). Thus insightsabout the costs or timing of abatement poli-cies derived from lAMs should be viewed astentative.

Goulder and Schneider (1999) consider theGDP losses from a carbon tax introduced in1995 and maintained at a constant rate of 25

Pacific and Asian Journal of Ener~y 1 Q( 1): 81-106


Thus, although actual numbers are uncer-tain, top-down models clearly find that strin-gent CO2 constraints are compatible with asignificantly increased material standard of Ii v-ing and they do not threaten to plunge the worldinto depression. This way of presenting mod-elling results (i.e. showing that the relativepaths of per capita GDP or consumption overthe 2151 century with and without carbon poli-cies are almost identical) deserves more atten-tion since there is a widespread impressionamongst policy makers and the general publicthat the opposite holds true.

ConclusionsNon-linearities and the likelihood of rapid,unanticipated events (surprises) require thatcosting methods use a wide range of estimatesfor key parameters or structural formulations,and that, when possible, results be cast inprobabilistic terms rather than central tenden-cies since the latter mask the policy-relevantwide range of potential results such a diversityof approaches implies. Costs need also to bepresented in more numeraires than just mon-etary ones. Because monetary cost estimatesmay more conceal than highlight the ethicaland moral dimensions of the potential climaticimpacts, in particular impacts on humanhealth, distribution of costs, or ecosystems. Theunderlying structural assumptions and param-eter ranges should be explicitly given in cost-ing analyses, to make the conclusions astransparent as possible. For example, while itis often acknowledged that a wide range of un-certainty accompanies estimates of climatedamages from scenarios of anthropogenic cli-matic change (owing to uncertainties in adap-tation capacity, synergistic impacts, etc.), it isless common (Moss and Schneider 1997) tohave a comparably wide set of estimates formitigation costs of carbon policies (e.g. a car-bon tax being a common analytic benchmark).Yet, the tighter range of mitigation cost esti-mates occurs in part because standard costingmethods make common assumptions about the

However, a different picture emerges fromanother perspective. This cost only has a mi-nor impact on the overall growth rates and in-come levels in the economy in the models usedto estimate it. In a survey of top-down studies,global per capita income by 2100 is assumedto be 5.4 times higher than at present if nocarbon abatement occurs. If carbon emissionsare kept at two-thirds of the present level forthe 2151 century, per capita income would be5.1 times higher (Azar 1996). Given assumedgrowth rates, the global income would be de-layed a couple of years before the higher in-come level is attained. Schneider (1993b),Grubb, Edmonds, ten Brink et al. (1993), andAnderson and Bird (1992) report similar ob-servations. There is near consensus, evenamong top-down modellers, regarding this.Note that the full range of potential environ-mental benefits from reducing the emissionshave not been included in these estimates (e.g.as Roughgarden and Schneider [1999] showed,a wide distribution of damage costs producesa very wide distribution of optimal carbontaxes).

The salience of the information providedby these modelling exercises to policy makersdepends on the political context. The threat ofclimate change is increasingly being recognizedas one of the most important challenges forthe 2151 century. There is mounting pressurefrom scientists and many different stakeholdergroups to take action to reduce emissions, butthe speed of action is still fairly low. Some poli-ticians and representatives from certain busi-ness sectors continue to oppose measures toreduce CO2 emissions. Perhaps, even moreimportantly, there is a genuine public concernthat emission reduction might reduce thematerial standard of living (in absoluteterms), force people into unemployment, or inthe words of President Bush during theUNCED (United Nations Conference onEnvironment and Development) meeting inRio de Janeiro in 1992, 'threaten the Ameri-can way of life'.



rapid changes in emissions triggering non-lin-ear climatic changes with potentially signifi-cant implications for costing.

In short, the key for authors of scientificassessments is transparency of assumptionsand the use of as wide a range of eventualities(and their attendant probabilities) as possibleto help decision makers become aware of thearguments for flexibility of policy options.

Acknowledgements

lack of pre-existing market failures or do notexplicitly account for the possibility of climatepolicy-induced technological changes reduc-ing mitigation cost estimates (Repetto andAustin 1997, Grubb, Ha-Duong, and Chapuis1994, Azar 1996, Goulder and Schneider1999). Although such endogenous growthformulations are controversial, cost estimatesmade in their absence need to be labelled asnot remotely covering the full range of pIa us i-ble values to prevent policy makers from mis-understanding the limited scope of the moreconventional results.

Moreover, in view of the wide range ofplau-sible climatic change scenarios available in theliterature - including a growing number ofrapid non-linear change projections - it is im-portant for costing analyses to consider manysuch scenarios, including the implications of

Table 3 Percentage differences between corn yields simulated with baseline observed climate (1984-93)and corn yields simulated with 1/3,2/3, and 3/3 of 2xCO2 climate change for three levels of adaptation:(1) no adaptation ('dumb farmer'), (2) perfect adaptation ('clairvoyant farmer'), and (3) adapatationlagged 20 years behind climate changes ('realistic farmer') .

A Central Iowa

Climate Change

(RegCM)

Perfect Adaptation Lagged Adaptation

(%) (%)

No Adaptation

(%)

-3

2

-3

-1

1/3 of2XCO22/3 of 2XCO23/3 of 2XCO2Mean of Thirds

-3

-8

-17

-10

22-0.31

B South Central Minnesota

Climate Change No Adaptation

(RegCM) (%)

Perfect Adaptation Lagged Adaptation

(%) (%)

1/3 of2XCO2

2/3 of 2XCO2

3/3 of 2XCO2

Mean of Thirds

8121010

13232420

8172216

Source Schneider, Easterling, and Mearns (2000)


Stephen H Schneider and Kristin Kuntz-Duriseti acknowledge partial support fromThePew Charitable Trusts (through the Pew Centeron Global Climate Change) andTheWinslowFoundation, Washington, DC. Christian Azaracknowledges financial support from the Swed-ish Council for Planning and Coordination ofResearch (Stockholm).


Azar C and Dowlatabadi H. 1999A review of technical change in assessmentsof climate change policyAnnual Review of Energy and the Environment 24:513-544.

Azar C and Johansson O. 1996Uncertainty and climate change-or theeconomics of twin peaksPaper resen ted at the A nnual conference of Euro-pean Environmental and Resource Economists,Lisbon, 1996.

Azar C and SternerT. 1996Discounting and distributional considera-tions in the context of climate changeEcological Economics 19:2169-185

Berk RA and Schulman D. 1995Public perceptions of global warmingClimatic Change 29(1): 1-33

Broome J. 1992Counting the Cost of Global ChangeCambridge, UK: The White Horse Press, 147pp.

CarterTR, Parry M L, HaraS'awa H, Nishioka S.1994IPCC Technical Guidelines for AssessingClimate Change Impacts and AdaptationsLondon, UK: University College 59pp.Chapman D, Suri V, and Hall S G. 1995 .

Rolling dice for the future of the planetContemporary Economic Policy 13(3): 1-9

ClineW.1992The Economics of Global WarmingWashington, DC: Institute of InternationalEconomics, 399pp.

Daily G C. 1997Nature's Services: Societal Dependence onNatural EcosystemsWashington, DC: Island Press, 392pp.

Ehrlich P R, Ehrlich A H, and Daily G. 1995The Stork and the PlowNew York: Putnam, 364pp.

Eriksson K E. 1994On discount, temporal justice, sustainabi1ityand limited foresightRicherche Economiche 48: 341-355

Fankhauser S. 1995Valuing Climate Change: The Economics of

the GreenhouseLondon: Earthscan, 180pp.

ReferencesAinslie G. 1991Derivation of "rational" economic behaviourfrom hyperbolic discount curvesAmerican Economic Association Papers and Proceed-ings 81(2): 334-340

Alexander S E, Schneider S H, and Lagerquist K.1997The interaction of climate and lifeIn Nature's Services: Societal Dependence on NaturalEcos_vstems, pp. 71-92, edited by G C DailyWashington, D C: Island Press, 392pp.

Anderson D and Bird C D. 1992Carbon accumulations and technicalprogress-a simulation of costsOxford Bulletin of Economics and Statistics 54: 1-29

Arrow K J, Cline W, Maler K G, Munasinghe M,Squitieri R, Stiglitz J. 1996Intertemporal equity, discounting andeconomic efficiencyIn Climate Change 1995: Economic and SocialDimensions of Climate Change- Second Assessmentof the Intergovernmental Panel on Climate Change,pp 125-144, edited by J P Bruce, H Lee, and E FHaitesCambridge, UK: Cambridge University Press447 pp.

Austin D. 1997Climate Protection Policies: Can We Affordto Delay?Washington D C:World Resources Institute, 38pp.

Ayres RU andWaltersJ. 1991The greenhouse effect: damages, costs andabatementEnvironmental and Resource Economics I: 237-270

Azar C. 1996Technological change and the long-run costof reducing CO2 emissionsFrance: INSEAD Centre for the management ofenvironment resources, working paper 96/84, 54pp.

Azar C. 1998The timing of CO2-emissions reduction-thedebate revisitedInternational Journal of Environment and Pollution10:3/4508-521

Azar C. 1999Weight factors in cost benefit analysis ofclimate changeEnvironmental and Resource Economics 13: 249-268



FankhauserS, Tol R S J; and Pearce D W. 1997The aggregation of climate change damages:a welfare theoretic approachEnvironnlenfal and Resource Economics 10:3 249-266

Goulder I~ H and Kennedy D. 1997Valuing ecosystems: philosophical bases andempirical methodsIn Nature's Sl!rvices: Societal Dependence on NaturalEcosystems, pp. 23-48, edited by G C DailyWashington, DC: Island Press, 392pp.

Goulder L H and Schneider S H. 1999Induced technological change and theattractiveness of CO2 emissions abatementpoliciesResource and Energy Econof1Zics 21: 211-253

Grubb M. 1997Energy technologies, systems and the timingof CO2 emissions abatementEnergy Policy 25(2): 159-72

Grubb M, Edmonds J, ten Brink P, Morrison M.1993The cost of limiting fossil-fuel CO2 emis-sions: a survey and an analysisAllIlua./ Rcview of Energy and the Environ1/lent 18:397-478

Grubb M, Ha-Duong M, and Chapuis T. 1994Optimizing climate change abatementresponses: on inertia and induced technologydevelopmentIn Integrative Assessment of Mitigation.. Impacts.. andAdaptation to Climate Change, pp. 513 - 534,edited by N Nakicenovic, W D Nordhaus, RRichels, F L TothLaxenburg, Austria: International Institute forApplied Systems Analysis, 669pp.

Ha-Duong M, Grubb M, and Hourcade J-C. 1997Influence of socio-economic inertia anduncertainty on optimal CO2 emission abate-mentNature 390(6657): 270-273

Harrod R. 1948Towards a Dynallmic EconomicsLondon: St. Martin's Press, 168pp.

Hasselmann K, Hasselmann S, Giering R, OcanaV, yon Storch H. 1997Sensitivity study of optimal CO2 emissionpaths using a simplified structural inte-grated assessment model (SIAM)Clinlatic Change 37(2): 345-386

Haywood J M, Stouffer RJ, Wetherald R T,Manabe S, RamaswamyV. 1997Transient response of a coupled model toestimated changes in greenhouse gas andsulfate concentrationsGeophysical Research Letters 24(11): 1335-1338

Heal G. 1997Discounting and climate change [an editorialessay]Climatic Change 37(2): 335~343

Intergovernmental Panel on Climate Change(IPCC). 1996aClimate Change 1995: The Science of ClimateChange- Contribution of Working Group Ito the Second Assessment Report of theIntergovernmental Panel on ClimateChange, edited by JT Houghton, L G MeiraFilho, B A Callander, N Harris, AKattenberg, K MaskellCambridge, UK: Cambridge University Press572pp.

Intergovernmental Panel on Climate Change(IPCC). 1996bClimate Change 1995: Impacts, Adaptationsand Mitigation of Climate Change:Scien-tific- Technical Analyses-Contribution ofWorking Group II to the Second AssessmentReport of tile Intergovernmental Panel onClimate Change, edited by R T Watson, M CZinyowera, and R H MossCambridge, UK: Cambridge University Press878pp.

Intergovernmental Panel on Climate Change(IPCC). 1996cClimate Change 1995: Economic and SocialD~'mensions of Climate Change-Contribu-tion of Working Group III to the SecondAssessment Report of the IntergovernmentalPanel on Climate Change, edited by, J PBruce, H Lee, and E F HaitesCambridge UK: Cambridge University Press448pp.

Intergovernmental Panel on Climatic Change(IPCC). Third Assessment Report, WorkingGroup II, first order draft, Cambridge UniversityPress.

Johansson-Stenman O. 2000On the value of life in rich and poor coun-tries and distributional weights beyondutilitarianismEnvironmental and Resource Economics (forthcoming)


104 S H Schneider. K Kuntz-Duriseti. C Azar

Mendelsohn R, Morrison W, Schlesinger M,Andronova N. 2000Country-specific market impacts of climatechangeClimatic Change 45(4-5): 553-569

Mendelsohn R, Nordhaus W, and Shaw D. 1996Climate impacts on aggregate farm value:accounting for adaptationAgricultural and Forest Meteorology 80: 1 55-66

Morgan G and Dowlatabadi H. 1996Learning from integrated assessment ofclimate changeClimatic Change 34(3-4): 337-68

Morgan M G and Keith D W. 1995Subjective judgments by climate expertsEnvironmental Science and Technology 29( I 0):468A-476A

Moss R H and Schneider S H. 1997Characterizing and communicating scien-tific uncertainty: building on the IPCCsecond assessmentIn Elements of Change, pp.90-135, edited by S JHassol and J KatzenbergerAspen, Colorado: AGCI (Aspen Global ChangeInstitute) 267pp .

Moss R H and Schneider S H. 1999Uncertainties in the IPCC TAR: Recommenda-tions To Lead Authors For More ConsistentAssessment and Reporting.(Paper presented at Conference of Inter Govern-mental Panel on Climate Change, Tokyo, 29 June-I July)

Munasinghe M. 1999Development, Equity and Sustainability(DES) in the Context of Climate Change,(Paper presented at Conference of Inter Govern-mental Panel on Climate Change, Tokyo, 29 June-1 July)

Kaiser H M, Riha S, Wilks D, Rossiter D,Sampath R. 1993A farm-level analysis of economic andagronomic impacts of gradual climatewarmingAmerican Journal of Agricultural Economics 75(2):387-398

Kaufmann R K. 1997Assessing the DICE model: uncertaintyassociated with the emission and retention ofgreenhouse gasesClimatic Change 33(4): 139-143

Khanna N and Chapman D. 1996Time preference, abatement costs, andinternational climate policy: an appraisal ofIPCC 1995Contemporary Economic Policy 14(2); 56-65

Kolstad C. 1993Looking vs. leaping: the timing of CO z controlin the face of uncertainty and learningIn Costs, Impacts, and Benefits of CO2 Mitigation,63-82" edited byY Kaya LaxenburgAustria: International Institute for SystemsAnalysis 595pp.

Kolstad C D, Kelly D L, and Mitchell G. 1999Adjustment costs from environmentalchange induced by incomplete informationand learningSanta Barbara, California: Department ofEconomics, University of California at SantaBarbara working paper, 44pp.

Manabe S and Stouffer R J. 1993Century scale-effects of increased atmos-pheric COz on the ocean-atmosphere systemNature 364(6434): 215-218

Manne A S, Mendelsohn R, and Richels R G.1995MERGE: a model for evaluating regional andglobal effects of GHG reduction policiesEnergy Policy 23( I): 17-34

Manne M and Richels R. 1997On stabilizing COz concentrations-cost-effective emission reduction strategies.Environmental Modeling and Assessment 2: 251-265

Markandya A. 1999The general framework for cost assessmentfor climate change.(Paper presented at Conference of Inter Govern-mental Panel on Climate Change, Tokyo, 29 June-I July)

National Academy of Sciences. 1991Policy Implications of Greenhouse WarmingWashington, DC: National Academy Press,127pp.

Nordhaus W D. 1990Count before you leapThe Economist (7 July): 19-22

Nordhaus W D. 1992An optimal transition path for controllinggreenhouse gasesScience 258(5086): 1315-1319


Costing non-linearities, surprises, and irreversible events 105

RootT L and Schneider S H. 1993Can large-scale climatic models be linkedwith mu1tiscale ecological studies?Conservation Biology 7(2): 256-270

Rosenberg N J (ed.). 1993Towards an integrated impact assessment ofclimate change: the MINK studyClimatic Change 24(1-2): 1-173

Rosenberg N J and Scott M J. 1994Implications of policies to prevent climatechange for future food securityGlobal Environmental Change 4: 49-62.

Rosenzweig C M, Parry M L, and Fischer G. 1994Potential impact of climate change on worldfood supplyNature 367(6459): 133-138

Roughgarden T and Schneider S H. 1999Climate change policy: quantifying uncer-tainties for damages and optimal carbontaxesEnergy Policy 27(7): 415-429

Schneider S H. 1993Pondering Greenhouse PolicyScience 259: 51001381

Schneider S H. 1996The future of climate: potential for interac-tion and surprisesIn Climate Change and World Food Security, 77-113, edited by T E DowningNATO ASI Series, Springer-Verlag, Heidelberg.

Schneider S H. 1997aLaboratory Earth: The Planetary Gamble WeCan't Afford to LoseNew York: Basic Books, 174pp.

Schneider S H. 1997bIntegrated assessment modeling of globalclimate change: transparent rational tool forpolicy making or opaque screen hidingvalue-laden assumptions?Environmental Modeling and Assessment 2 (4):229-248

Schneider S H, Easterling WE, and Mearns L O.2000Adaptation: essential to impact assessmentbut still highly subjectiveClimatic Change 45(1): 203-221

Schneider SHand Goulder L H. 1997Achieving low-cost emissions targetsNature 389(6646): 13-14

Nordhaus W D. 1994aExpert opinion on climatic changeAmerican Scientist 82(1): 45-52

Nordhaus W D~1994b.Managing the Global Commons: the Eco-nomics of Climate Change.Cambridge, and Massachusetts: MIT Press 213pp.

Nordhaus W D. 1997Discounting in economics and climatechangeEditorial Climatic Change 37(2): 315-328

Peck S C and Teisberg T J. 1992CETA: a model of carbon emissions trajec-tory assessmentThe Energy Journal 13(1): 55-77

Price C. 1993Time, Discounting and ValueOxford, UK: Blackwell publishers, 393pp.

Rabl A. 1996Discounting of long term costs: what wouldfuture generations want us to do?Ecological Economics 17:3137-145

Rahmstorf S. 1999Shifting seas in the greenhouse?Nature 399:6736523-524

Ramsey F P. 1928A mathematical theory of savingThe Economic Journal 138: 543-549

Rawls J. 1972A Theory of JusticeOxford University Press, UK, 538pp.

Reilly J, Baethgen W, Chege F E, van de Geijn SC, Erda L, Iglesias A, Kenny G, Patterson D,Rogasik J, Rotter R, et al. 1996Agriculture in a changing climate: impactsand adaptationIn Climate Change 1995: Impacts, Adaptationsand Mitigation of Climate Change-Scientific-Technical Analyses (Contribution of WorkingGroup II to the Second Assessment Report of theIntergovernmental Panel on Climate Change),427-67, edited by R TWatson, M C Zinyowera,and R H MossCambridge, UK: Cambridge University Press878pp.

Repetto R and Austin D. 1997The Costs of Climate Protection: a Guide forthe PerplexedWashington, DC:World Resources Institute, 51pp.



Titus J and Narayanan V. 1996The risk of sea level rise: a Delphic MonteCarlo analysis in which twenty researchersspecify subjective probability distributionsfor model coefficients within their respectiveareas of expertiseClimatic Change 33(2): 151-212

West J J and Dowlatabadi H. 2000Storms, investor decisions and the economicimpacts of sea level rise.Climatic Change (in press)

WigleyT, Richels R, and Edmonds J. 1996Economics and environmental choices in thestabilisation of atmospheric CO2 concentra-tionsNature 379(6562): 240-243

Yohe G. 1989The cost of not holding back the sea-eco-nomic vulnerabilityOcean and Shoreline Management 15: 233-255

Yohe G. 1991Uncertainty, climate change, and the eco-nomic value of informationPolicy Science 24: 245-269

Yohe G, Neumann J, Marshall P, Ameden H.1996The economic cost of greenhouse inducedsea level rise for developed property in the'United StatesClimatic Change 32(4): 387-410

Yohe G and Wallace R. 1996Near-term mitigation policy for globalchange under uncertainty: minimizing theexpected cost of meeting unknown concen-tration thresholdsEnvironmental Modeling and Assessment 2: 47-68

Schneider SHand Thompson S L. 1985Future changes in the atmosphereIn The Global Possible, 397 -430, edited by R RepettoNew Haven, Connecticut:Yale University Press,538pp.

Schneider S H, Turner B L, and MorehouseGarriga H. 1998Imaginable surprise in global change scienceJournal of Risk Research 1 (2): 165-185

Schultz P A and Kasting J F. 1997Optimal reductions in COz-emissionsEnergy Policy 25(5): 491-500

Sidgwick H. 1890The Methods of Ethics, 4th editionLondon: Macmillan, 522pp.

Smith J and Tirpak D. 1990The potential effects of global climate changeon the United States.New York: Hemisphere Publishing Corporation,689pp.

Solow R. 1992An almost practical step towardssustainability[An invited lecture on the occasion of the fortiethanniversary of Resources for the Future, 8October, Washington]

Spash C Land d' Arge R C. 1989The greenhouse effect and intergenerationaltransfersEnergy Policy 17: 88-96

StockerT F and Schmittner A. 1997Influence ofCOz emission rates on thestability of the thermohaline circulationNature 388(6645): 862-864

Thompson S L and Schneider S H. 1982CO2 and climate: the importance of realisticgeography in estimating the transientresponseScience 217: 1031-33

Pacific and Asian Journal of Energy 10(1): 81-10~

Date post:	11-Feb-2022
Category:	Documents
Upload:	others
View:	3 times
Download:	0 times

Costing non-linearities, surprises, and irreversible events

Documents