18.4.2 Consideration of costs and damages avoided and/or benefits gained
Various approaches have been taken since the TAR to estimate the size of climate change damages that can be avoided by emissions reduction. Among the global integrated assessments reviewed in the previous sub-section, cost-effectiveness models (by far the most widely used decision analysis framework) do not include impacts, hence they cannot measure avoided damages either. In contrast, CBAs of greenhouse-gas emissions reduction (e.g., Nordhaus, 2001) necessarily estimate the avoided damages of climate change but rarely report them. Economic assessments of marginal damage costs (e.g., the incremental impact of an additional tonne of carbon emissions) provide a means of comparing damages avoided with marginal abatement costs. Such studies typically cover a range of sectors and report damage functions and estimates for scenarios of climate change, and increasingly reference scenarios of socio-economic vulnerability.
Tol (2005b) reviewed the avoided-damage literature, including 103 estimates from 28 papers published from 1991 to 2003. Some of the reviewed estimates include only a few impacts; other estimates include a wide range of impacts, including low-probability/high-impact scenarios (see Chapter 20 for further discussion). Tol (2005b) finds that most studies (72% when quality-weighted) point to a marginal damage cost of less than US$50 per tonne carbon (/tC). He also finds a systematic, upward bias in the grey literature. For instance, the 95th percentile falls from US$350/tC to US$245/tC if estimates that were not peer-reviewed are excluded. For a 5% discount rate, a value used by many governments (Evans and Sezer, 2004), the median estimate is only US$7/tC; for a 3% discount rate, it is US$33/tC.
Downing et al. (2005) updated the Tol (2005b) analysis to a 2005 base year: the very likely range of estimates runs from -US$10 to +US$350/tC; peer-reviewed estimates have a mean value of US$43/tC with a standard deviation of US$83/tC. Incorporating results from FUND (2005 version) and PAGE2002, Downing et al. (2005) find that £35/tC (at year 2000 values, or US$56/tC) is a credible lower benchmark for the social cost of carbon (as identified by the UK Government in Clarkson and Deyes, 2002). In FUND, with the Green Book discounting scheme and equity weighting, there is about a 40% chance that the social cost of carbon exceeds £35/tC. Estimates of the central tendency (whether the average or median) or upper benchmark were not agreed in that assessment, due to the limitations in our knowledge of climate impacts and the critical role of the decision perspective (see Section 18.5).
Stern (2007), including a higher level of risk of adverse impacts that are poorly represented in existing models and accepting a public policy framework that includes low discounting of the future, reports a social cost of carbon of US$304/tC (US$85/tCO2, at pounds sterling 2005 values) from the PAGE2002 model. The range of estimates is quite large and Stern (2007) acknowledges that his central estimate is higher than most studies and is “keenly aware of the sensitivity of estimates to the assumptions that are made”.
Note that the estimates of avoided damages are highly uncertain. A survey of fourteen experts in estimating the social cost of carbon rated their estimates as low confidence, due to the many gaps in the coverage of impacts and valuation studies, uncertainties in projected climate change, choices in the decision framework and the applied discount rate (Downing et al., 2005).
The marginal damage cost only gives the value of the last unit of the damage avoided, not the total avoided damage, which is seldom estimated (see the literature review and papers in Corfee-Morlot and Agrawala, 2004). Nonetheless, as a first approximation of the avoided damages, one should multiply the tonnes of carbon emissions reduced by the marginal damage cost.
Several studies have attempted to calculate total economic damages from disparate impact studies. Warren (2006) reports a long list of ecosystem impacts at 2°C warming and below, billions of people at risk from water stress (without adaptation) and political tension in Russia. As the impact estimates are taken from different studies, with different models and different scenarios, this method introduces additional uncertainties: the difference in impact may be due to different warming scenarios, but also due to differences in models, data, economic scenarios and even subject and area of study. Furthermore, it is difficult to compare how impacts change with additional degrees of climate change, although the work does suggest that there are an increasing number of negative impacts at higher temperatures. Warren’s (2006) study is often qualitative and it is unclear whether the studies are representative of the literature (or the population of affected sectors), or whether adaptation is included. On avoidable damage, this study paints a bleak picture. At 2°C warming, which may be difficult to avoid, 97% of coral reefs and 100% of Arctic sea ice would be lost. Avoided damage is therefore less than 3% of coral reefs, and no Arctic sea ice. Hare (2006) also offers impact estimates for various warming scenarios, with the same limitations as for Warren (2006). Hitz and Smith (2004) review damage functions related to global mean temperature but do not aggregate to overall damages. Arnell et al. (2002) and Parry et al. (2004) use internally consistent models and scenarios, and report numbers for avoided damages, measured in millions of people at risk. Water resources and malaria dominate their results, but the underlying models do not account for adaptation and keep socio-economic development at 1990 levels, although populations grow.
Relatively few studies have documented damages avoided in terms of specific mitigation scenarios. Bakkenes et al. (2006) study the implications of different stabilisation scenarios on European plant diversity. Mitigation is not considered, even though biofuels and carbon plantations would substantially affect vegetation. Under the A1B scenario, plants would lose on average 29% of their current habitat by 2100, with a range between species from 10% to 53%. Stabilisation at 650 ppm would limit this to 22% (6-42%), and at 550 ppm to 18% (5-37%). With unmitigated climate change, nine plant species would disappear from Europe, but eight new ones would appear. Stabilisation would limit the number of plant disappearances from nine to eight species. In all five studies, adaptation (except in some parts of the Parry study) and the effects of mitigation on impacts are not included (see Section 18.4.1). Nicholls and Lowe (2004) estimate the avoided impact of sea-level rise due to mitigation. Because sea level responds so slowly to global warming, avoided impacts are small, at least over the 21st century. Nicholls and Lowe (2004) ignore the costs of emissions reduction; Tol (2007) shows that the bias is negligible for coastal-zone impacts. Nicholls and Lowe (2004, 2006) argue that adaptation and mitigation should be applied together for coastal zones, with mitigation to minimise the future commitment to sea-level rise and adaptation to adapt to the inevitable changes. Nicholls and Tol (2006) and Nicholls et al. (2007) also explore the economic impacts of sea-level rise.
Tol and Yohe (2006), using the integrated assessment model, Climate Framework for Uncertainty, Negotiation and Distribution (FUND), conclude that the most serious impacts of climate change can be avoided at an 850 ppm CO2-equivalent stabilisation target for greenhouse-gas concentrations, and that incrementally avoided damages get smaller and smaller as one moves to more stringent stabilisation targets. For a 450 ppm CO2-equivalent stabilisation target, climate-change impacts may actually increase as the reduction of sulphur emissions may lead to warming and as abatement costs slow growth and increase vulnerability. However, FUND includes a wide range but not all impacts, represents impacts in a reduced form, does not capture discontinuities or interactions between impacts, models climate change as being smooth, and does not include the ancillary benefits of reductions in sulphur. Other models also find that climate policy would reduce sulphur emissions to levels below what is required for acidification policy (e.g., Van Vuuren et al., 2006). Other integrated assessment models have yet to produce comparable analyses.
Abatement may, but need not, reduce the probability of extreme climate scenarios, such as a shut-down of the thermohaline circulation (Gregory et al., 2005) and a collapse of the West Antarctic ice sheet (Vaughan and Spouge, 2002). The few studies on the effects of drastic sea-level rise show large impacts (Schneider and Chen, 1980; Nicholls et al., 2005; Tol et al., 2006) but opinions on the impacts of a thermohaline circulation shut-down are divided (Rahmstorf, 2000; Link and Tol, 2004).
Additional assessments of damages avoided by mitigation are also provided in other chapters of this report. Chapter 20 finds that estimates of the social cost of carbon expand over at least three orders of magnitude and notes that globally aggregated figures are likely to underestimate the full costs, masking differences in impacts across sectors and regions/countries. It concludes that “it is very likely that climate change will result in net costs into the future, aggregated across the globe and discounted to today; it is very likely that these costs will grow over time”. The WGIII AR4 in Chapter 3 (Fisher et al., 2007) observes that most (but not all) analyses which use monetisation suggest that social costs of carbon are positive, but the range of values is wide and is strongly dependent on modelling methodology, value judgements and assumptions. It concludes that large uncertainties persist, related to the cost of mitigation, the efficacy of adaptation, and the extent to which the negative impacts of climate change, including those related to rate of change, can be avoided. See Box 18.2 for a summary of the WGIII AR4 conclusions on damages avoided with different stabilisation scenarios.
Box 18.2. Analysis of stabilisation scenarios
The WGIII AR4, in Chapter 3 (Section 3.5.2), looks across findings of the WGI and WGII AR4 to relate the long-term emissions scenarios literature to climate-change impact risks at different levels of global mean temperature change based on key vulnerabilities (as defined in Chapter 19). It builds on the WGI AR4 findings, which outline the probabilities of exceeding various global mean temperatures at different concentration levels (Tables 3.9 and 3.10 in Fisher et al., 2007). The relationships are based on a key finding of the WGI AR4 that there is at least an 83% probability for climate sensitivity to be at or below 4.5°C, while the best estimate is for climate sensitivity to be 3°C. The WGIII AR4 organises the stabilisation scenarios literature by the level of stringency of the scenario, setting out six groups (I-VI) that cover the full range of more to less stringent global warming objectives, in the form of concentrations (ppm) or radiative forcing (W/m2). Table 3.9 uses the WGI AR4 findings to relate increases in global mean temperature to concentration targets, while Table 3.10 relates these outcomes to the emissions pathways associated with alternative stabilisation scenarios. (An important caveat is that these relationships do not consider possible additional CO2 and CH4 releases from Earth-system feedbacks and thus may underestimate required emissions reductions.)
Regarding climate-change impact risks and key vulnerabilities, this literature is organised around increase in global mean temperature. Chapter 19 shows that the following benefits would accrue from constraining temperature rise to 2°C above 1990:
- lowering the risk of widespread deglaciation of the Greenland ice sheet**;
- avoiding large-scale transformation of ecosystems and degradation of coral reefs***;
- preventing terrestrial vegetation becoming a carbon source*/**, constraining species extinction to between 10% and 40%*, and preserving many unique habitats (see Chapter 4, Table 4.1 and Figure 4.5);
- preventing flooding, drought and water-quality declines***, global net declines in food production*/•, and more intense fires**.
Other benefits of this constraint include reducing the risks of extreme weather events**, and of at least partial deglaciation of the West Antarctic ice sheet (WAIS)* (see Chapter 19, Section 19.3.7). By comparison, constraining temperature change to not more than 3°C above 1990 levels will still avoid commitment to widespread deglaciation of the WAIS* and commitment to possible shut-down of the Meridional Overturning Circulation/• but results in significantly lower avoided risks and impacts in most other areas (Chapter 19, Section 19.3.7).
(Confidence ratings are as provided by WGII Chapter 19 authors: /• = low confidence, * = medium, ** = high, and *** = very high confidence.)
Overall, there are only a few studies that estimate the avoided impacts of climate change by emissions reduction. Some of these studies ignore adaptation and mitigation costs. Many published studies of damages in sectors that are quantified in economic models (but mostly market-based costs and related to incremental projections of temperature) and with discount rates commonly used in economic decision-making (e.g., 3% or higher) lead to low estimates of the social cost of carbon. In general, confidence in these estimates is low. The paucity of evidence is disappointing, as avoiding impacts is presumably a major aim of climate policy. CBAs of climate change implicitly estimate avoided damages and suggest that these do not warrant very stringent emissions reduction (see Section 18.4.1). Similarly, although ecosystem impacts may be large, avoidable impacts may be much smaller. With few high-quality studies, confidence in these findings is low. This is a clear research priority. The use of the social cost of carbon in decision-making on mitigation also warrants further exploration.