2

Framework for Estimating the Social Cost of Carbon

This chapter provides an overview of the steps involved in estimating the social cost of carbon and the committee’s recommendations for how they should be organized in future updates. The committee discusses how uncertainty might be characterized in such a framework and the level of geographic, sectoral, and temporal detail involved. The frequency of updates to SC-CO₂ estimates and how the process of updating SC-CO₂ estimates might be structured is also discussed.

STRUCTURE OF THE ESTIMATION PROCESS

Estimating the SC-CO₂ involves four steps: (1) projecting future global and regional population, output, and emissions; (2) calculating the effect of emissions on temperature, sea level, and other climate variables; (3) estimating (explicitly or implicitly) the physical impacts of climate and, to the extent possible, monetizing those impacts on human welfare (i.e., estimating net climate damages); and (4) discounting monetary damages to the year of emission.

The committee structured its work, conclusions, and recommendations around these four parts of a framework—socioeconomic factors and emissions, climate, impacts and damages, and discounting—which are characterized as modules. Each of these modules is comprised of data, conceptual formulations and theory, computer models and other analysis frameworks. And, to some extent, each is supported by its own specialized disciplinary expertise. Estimation of the SC-CO₂ involves the

integration of these four modules, taking account when possible the interdependencies among them.

Studies supporting SC-CO₂ estimation have used integrated assessment models (IAMs) that incorporate some or all of the four components in a single model: Box 2-1 details some key terminology related to model-

ing used in this report. The IWG used three reduced-form IAMs—DICE (Dynamic Integrated Climate-Economy model), FUND (Framework for Uncertainty, Negotiation and Distribution model), and PAGE (Policy Analysis of the Greenhouse Effect model)—to compute the SC-CO₂, pooling the final SC-CO₂ estimates from each model at the end of the analysis. The essence of the committee’s approach is to unbundle the four steps of

the analysis. Rather than averaging the results from three separate SC-IAMs, the committee suggests a single framework with modules designed to capture uncertainty at each step. This report focuses on how each module of the analysis could be constructed.

Figure 2-1 illustrates the parts of the SC-CO₂ estimation process in terms of the committee’s recommended modular framework, showing how information is exchanged among the modules, leading ultimately to the SC-CO₂ estimate. The flow of intermediate results in the framework is shown in solid lines. The dashed lines introduce additional possible interactions among components of the estimation that are discussed below.

As Figure 2-1 shows, a socioeconomic module (detailed in Chapter 3) generates projections of greenhouse gas emissions for input to the climate module, as well as estimates of future population and gross domestic product (GDP) that are direct inputs to the damages module and the discounting module (Box 2-2 details the key economic and related terms used in this report). The projected emissions paths serve as a baseline to which an emission pulse is added in order to represent the incremental impact of an additional ton of CO₂ released in a particular year.

Based on projected emissions, a climate module (detailed in Chapter 4) generates estimates of greenhouse gas concentrations, temperature change, sea level rise, and other needed climate variables. Along with population and GDP, these climate results are then inputs into a damages module (detailed in Chapter 5) that calculates the physical impacts of climate variables on outcomes that affect human welfare and, when possible, monetizes on a year-by-year basis the net damages caused by the climate change due to CO₂ emissions. The grey dashed outline around the damages module in Figure 2-1 indicates that regional or sectoral socioeconomic and climate data will likely be necessary either as direct inputs to impact functions or for their calibration. The figure also shows that non-monetized impacts may also be included in representations of the cost of CO₂ emissions, albeit in physical rather than monetary terms.

The purpose of a discounting module (detailed in Chapter 6) is to integrate the future stream of monetized damage estimates into a single present value for each state of the world generated by the earlier steps of the SC-CO₂ estimation process. The committee suggests an approach yielding three discount rate scenarios, with each scenario having a distribution of SC-CO₂ values that is determined by all the other sources of uncertainty incorporated in the SC-CO₂ estimation process. To date, the IWG has focused on scenarios with fixed discount rates of 2.5 percent, 3 percent, and 5 percent. The dotted line in Figure 2-1 shows GDP and population as recommended future inputs into the discounting module to capture the relationship between the year-to-year discount rate and growth in per capita GDP.

AN INTEGRATED, MODULAR FRAMEWORK

Current estimates of the SC-CO₂ are obtained by averaging estimates of monetized damages produced by the three SC-IAMs, each of which contains its own climate component and set of damage functions. Although a common set of socioeconomic scenarios and a common distribution of

equilibrium climate sensitivity¹ (ECS) is applied to each SC-IAM, their climate and damage components differ significantly. The final distribution over the SC-CO₂ estimates is based on the average and range of the different components of these structurally distinct models (see Figure 1-1, in Chapter 1).

A previous study has documented the differences in the assumptions and functional forms embedded in the climate components of the three SC-IAMs (Rose et al., 2014b). Of particular concern to the committee is that, even under a common value of the ECS, significant differences in climate modeling structure and climatic response underlie the estimates from the three SC-IAMs that are being averaged. Such differences would be informative if they systematically represented structural uncertainty in the climate system—that is, uncertainty about which is the correct modeling structure to use—but in practice, and as discussed in more detail in Chapter 4, the differences arise instead from uncoordinated modeling choices of the individual model developers.

Because of these differences, in its Phase 1 report (National Academies of Science, Engineering, and Medicine, 2016) the committee suggested that the IWG undertake efforts to adopt or develop a common climate module with three characteristics:

A similar argument for a common module can be made for the socioeconomic and impact/damage components of the analysis, in effect indicating an approach to the IWG’s task that places heavy emphasis on improving the scientific and information basis of each of the four main components of the analysis. The committee contends that this modular approach is superior to averaging the SC-CO₂ estimates from separate IAMs that may depend on inconsistent assumptions or on assumptions that, when averaged, do not yield an overall distribution of uncertainty that is consistent with the best available evidence.

__________________

¹ ECS measures the long-term response of global mean temperature to a fixed forcing, conventionally taken as an instantaneous doubling of CO₂ concentrations from their preindustrial levels. See Chapter 4 for further discussion.

Such a modular estimation framework can help ensure consistency of key assumptions and can aid in the rigorous and transparent characterization of uncertainty at each stage of the estimation process. It can also provide a means for transparently identifying the inputs, outputs, and linkages among the various stages of the SC-CO₂ estimation process. This modularity can thereby enable expert groups in the broad scientific community to evaluate aspects of the process that are within their disciplinary expertise, while ensuring that these elements are coherently integrated.

The main risk in a focus on individual modules is a failure to identify and take proper account of feedbacks and other interactions among components of the human-climate system that cut across these modular boundaries. This concern is addressed below, as well as potential future research activities that could be undertaken to address it.

The committee recognizes that models developed in academic research may require substantial modification before being appropriate for use in estimating the SC-CO₂, as the purpose of that research is not to generate estimates for use in regulatory impact analysis. The committee leaves to the discretion of the IWG the best way to assemble results from the scientific literature into a modular framework for estimating the SC-CO₂. This may involve issuing contracts to researchers outside of the U.S. government and/or choosing to have the analysis performed within the government.

Subsequent chapters outline criteria that are specific to each module: below is a general set of standards that apply to all analytical efforts to estimate the SC-CO₂.²

__________________

² The committee notes that the criteria listed in Recommendation 2-2 reinforce and are consistent with U.S. Office of Management and Budget (OMB) Circular A-4 guidance for regulatory impact analysis, which includes general guidelines for “Transparency and Reproducibility of Results” and the “Treatment of Uncertainty.”

FEEDBACKS AND INTERACTIONS

Over time, successful implementation of such a framework will require attention to the interactions among the modules, and necessitate adaptation of the overall structure to take advantage of ongoing research on the human-environment-climate system. One example of such interaction is suggested in Figure 2-1 (above): a dashed line indicates that climate damages, evaluated in the damages module, could feed back onto greenhouse gas emissions, as represented in a socioeconomic module. If the output of the damages module shows a reduction in GDP, for example, that reduction may affect the projected GDP in subsequent years and thus the projected emissions from the socioeconomic module. Similarly, the temperature- and CO₂-driven impacts on crop yields projected by the damages module may affect agricultural productivity in the socioeconomic module, and the warming-driven impacts on air-conditioner adoption and use could affect energy use and thus CO₂ emissions projected by the socioeconomics module. In the current SC-CO₂ estimation, the emissions projections are exogenous in all three SC-IAMs and GDP projections are exogenous in two of the three SC-IAMs. In the current framework, there is therefore little feedback from climate or damages to emissions and GDP.

Ongoing research on climate impacts/damages, integrated assessment, and Earth system modeling is revealing interactions among the components of SC-CO₂ estimation that go beyond the above examples of feedback from climate impacts/damages to socioeconomic projections, suggesting additional future changes in implementation of the four-module framework recommended in this report. For example, it is now recognized that some of the most severe impacts of climate change on particular regions in specific sectors result from interactions between them and impacts in other regions or sectors (Oppenheimer et al., 2014). These interactions can occur at the physical level: for example, if climate change causes a particular region to become hotter and dryer, it might increase the competition for limited water between agricultural, power plant cooling, and household and commercial uses (see Taheripour et al., 2013). These shortages could then make both food and energy more expensive in some regions, which could have general equilibrium effects on other economic sectors in that region or elsewhere (Baldos et al., 2014). Indeed, increases in temperature and decreases in precipitation are already occurring in a number of major growing regions, leading to the need for more irrigation,

which is largely being met by depleting ground water resources (Grogan et al., 2015). As ground water aquifers are drained, water becomes even more scarce, which may reduce agricultural production in some very vulnerable low-income regions (Zaveri et al., 2016). Other interactions may mitigate impacts, including, for example, the reduction of the health impacts of heat stress by increased air conditioning.

Such interactions are being explored in detailed-structure IAMs (Reilly et al., 2012a). One prominent example is the interaction between greenhouse gas mitigation and urban and regional air pollution policies (Reilly et al., 2007; Chuwah et al., 2013; Nam et al., 2014). Another is the study of competition for water for both agriculture and power plant cooling that will occur in hotter and dryer climates, as well as of the impacts on water and land (and the resulting land emissions) of global policies that rely on massive increases in biofuels (Reilly et al., 2012b; Daioglu et al., 2014; Rose et al., 2014a).

Although the literature on these feedbacks and interactions is advancing, many of the relevant studies consider these phenomena one at a time, perhaps missing interactions among them. Some yield results only in physical terms and do not proceed to economic measures, and the studies to date frequently consider only one country or region and do not provide a basis for extension to the world as a whole. Thus, opportunities for incorporating the relevant feedbacks and interactions in a modular approach depend on the state of scientific knowledge as it will emerge from ongoing research, as well as on details of the damage functions and the nature of the modeling frameworks used for the socioeconomics projections.

Even at the simplest level of interaction among modules, shown in Figure 2-1, there will be need for careful attention to the flow of information among them. And care will be required during the development phase to maintain consistency among modules. For example, components will require consistency or appropriate conversion across units of measurement, time steps, uncertainty representation, and regional and sectoral specification. Additionally, the modules may differ in their choice of software development systems. This task will grow more challenging with consideration of additional feedbacks and other interactions that will require stronger and tighter coupling of the modules. It will thus be desirable to choose an integration platform that can accommodate change in internal module structure and interactions; it may even be desirable to integrate the components into a single computational framework.

GEOGRAPHIC, SECTORAL, AND TEMPORAL DETAIL

Implementation of a modular approach will entail decisions about the level of regional and sectoral detail in each module and about the time horizon over which estimates are computed. These issues are discussed below, with specific focus on the United States.

Level of Geographic and Sectoral Detail

As the dashed box in Figure 2-1 indicates, estimation and implementation of damage functions may require climate and/or socioeconomic inputs at a regional and/or sectoral level. The level of regional and sector disaggregation necessary will be dictated by the level of disaggregation of

damages. The damage module may specify separate damage functions for different impact sectors (e.g., agriculture, sea level rise, electricity generation) and different regions (e.g., United States, India, China, sub-Saharan Africa). Disaggregation may also be necessary as an intermediary step toward calibration of an aggregate damage function. Chapter 3 discusses approaches for providing disaggregated population and GDP by region and sector, and Chapter 4 discusses methods for regional downscaling of temperature and sea level rise.

In the near term, probabilistic spatially disaggregated projections of population, GDP, temperature and sea level rise are particularly challenging. However, deterministic disaggregation (i.e., point estimates of variables) could be used as direct inputs to regional and sectoral damage functions or to calibrate global damage functions based on detailed regional and sectoral damage characterizations.

In the longer run, it may be possible to use models with regional and sectoral detail as the basis of socioeconomic projections and to provide probability distributions of disaggregated values for population and GDP. Similarly, it may be possible in the longer term to obtain probability distributions defined over spatially disaggregated climate variables.

U.S. Damages

Because CO₂ emissions have global impacts regardless of the country from which they originate, both research and IWG efforts to estimate the SC-CO₂ have focused on total global damages, rather than the damages to any individual country. In 2010 the Interagency Working Group on the Social Cost of Carbon (2010, p. 10) stated that “because of the distinctive nature of the climate change problem, we center our current attention on a global measure of SCC [SC-CO₂].” At the same time, the IWG recognized that this approach “represents a departure from past practices, which tended to put greater emphasis on a domestic measure of SCC (limited to impacts of climate change experienced within U.S. borders).” Nonetheless, the IWG asserted its flexibility, noting that (p. 10):

In its updates to the SC-CO₂, the IWG has consistently supported a focus on global values, as reflected in many Technical Support Document

updates (Interagency Working Group on the Social Cost of Carbon, 2010, pp. 10-11; 2013a, p. 14; 2013b, p. 14; 2015, p. 14):³

Despite this consistent focus, the IWG did explore the basis for estimating the SC-CO₂ for the United States (Interagency Working Group on the Social Cost of Carbon, 2010, p. 11): “[A]s an empirical matter, the development of a domestic SCC is greatly complicated by the relatively few region- or country-specific estimates of the SCC in the literature.” Using only the FUND model (which has regional disaggregation to its damage functions), the IWG noted an average U.S. benefit of about 7-10 percent of the global benefit across the scenarios it analyzed. Alternatively, the IWG found that “if the fraction of GDP lost due to climate change is assumed to be similar across countries, the domestic benefit would be proportional to the U.S. share of global GDP, which is currently about 23 percent.” On this basis, the IWG “determined that a range of values from 7 to 23 percent should be used to adjust the global SCC to calculate domestic effects,” recognizing that “these values are approximate, provisional, and highly speculative.” Nonetheless, as described in the IWG’s 2015 Response to Comments (Interagency Working Group on the Social Cost of Carbon, 2015b, p. 31), some commenters have asserted that domestic damage estimates have received inadequate attention.

Correctly calculating the portion of the SC-CO₂ that directly affects the United States involves more than examining the direct impacts of

__________________

³ The 2016 update of the Technical Support Document uses similar language on p.17 (Interagency Working Group on the Social Cost of Greenhouse Gases, 2016b) but also cites the worldwide commitment by many countries to reducing greenhouse gases in the signing of the Paris Agreement on April 22, 2016.

climate that occur within the country’s physical borders, which is what the 7-23 percent range is intended to capture. Climate damages to the United States cannot be accurately characterized without accounting for consequences outside U.S. borders. As the IWG noted (Interagency Working Group on the Social Cost of Carbon, 2010), climate change in other regions of the world could affect the United States through such pathways as global migration, economic destabilization, and political destabilization. In addition, the United States could be affected by changes in economic conditions of its trading partners: lower economic growth in other regions could reduce demand for U.S. exports, and lower productivity could increase the prices of U.S. imports. The current SC-IAMs do not fully account for these types of interactions among the United States and other nations or world regions in a manner that allows for the estimation of comprehensive impacts for the United States.

In addition, the United States may choose to use a global SC-CO₂ in order to leverage reciprocal measures by other countries (Kopp and Mignone, 2013; Howard and Schwartz, 2016). U.S. emissions impose most of their damage beyond U.S. borders, and climate damages to U.S. citizens will largely depend on emissions and mitigation choices by other countries. Climate damages and mitigation benefits to each country are thus determined by the global effort, and the potential to leverage foreign mitigation supports a domestic SC-CO₂ estimate augmented by the expected foreign leverage (Pizer et al., 2014). Considering all these factors, there are reasons to consider a global SC-CO₂ and what constitutes domestic impact in the case of a global pollutant.

Time Horizon

In concept, the SC-CO₂ assesses the total discounted damage to social welfare caused by an emission of CO₂ occurring in a particular year,

which results in damages that can persist several centuries into the future. Thus, the question arises of what time horizon should be used for the analysis. In the context of the socioeconomic, damage, and discounting assumptions, the time horizon needs to be long enough to capture the vast majority of the present value of damages.⁴ The length of this horizon depends on the rate at which undiscounted damages grow over time and on the rate at which they are discounted. Longer time horizons allow for representation and evaluation of longer-run geophysical system dynamics, such as sea level change and the carbon cycle; however, they involve greater speculation and uncertainty about socioeconomic conditions and emissions. It will be informative, for analytic transparency and decision making, for the IWG to report the share of the SC-CO₂ accruing over different time horizons. Such reporting would provide a sense of the relative importance of very long-term impacts to the overall estimate.

TREATMENT OF UNCERTAINTY

The inputs and the outputs of each module of the SC-CO₂ analysis have inherent uncertainties, as do the structures and parameters of the modules themselves. The future growth rates of population and output are uncertain. The emissions associated with any future GDP path depend on policies to control greenhouse gas emissions and on the evolution of energy technologies, energy markets, and land-use patterns—all of which are uncertain. Although the basic physics of the climate system are well established, the parameters linking emissions to mean global temperature and other climate variables are not known with certainty. Similarly, there is uncertainty in the translation of physical climate changes into impacts and damages. Given these numerous uncertainties inherent in SC-CO₂ estimation, the IWG necessarily needs to focus its analytical attention on incorporating the most important sources of uncertainty, rather than seeking to incorporate all possible sources of uncertainty.

Both parametric uncertainty and structural uncertainty are in each of the modules that comprise the SC-CO₂ framework. Parametric uncertainty is uncertainty about the value of various parameters in a model (or models) used in a module. A physical climate example of parametric uncertainty is uncertainty in the strength of known feedbacks that amplify or dampen the sensitivity of the global mean temperature to climate forcing. Structural uncertainty is uncertainty about what model constitutes the best framework for understanding what one wishes to project. A physical climate example of structural uncertainty is the possible presence

__________________

⁴ Vast majority” is a deliberately vague term that signals much more than 50 percent, but not 100 percent.

of unknown feedbacks, which may be hinted at by the geological record of past climate responses to climate forcing similar to the magnitude of recent anthropogenic forcing. Another example of structural uncertainty arises with respect to climate damages because of unknown damage pathways associated with abrupt climate change.

Approaches to Decision Making under Uncertainty

The standard approach to benefit-cost evaluation under uncertainty is expected value analysis of the consequences for social welfare, with additional consideration of the distribution of consequences around the expected value. As explained in Chapter 1, this is the regulatory approach that underlies regulatory impact analysis under Executive Order 12866, for which the SC-CO₂ was developed. Under this approach, one evaluates each regulation by the resulting change in expected welfare, that is, by the effect of that regulation on social welfare in each possible state of the world weighted by the probability of that state. The results of this approach depend on the probabilities associated with each state of the world.

Other approaches make use of multiple probability distributions over states of the world. One such approach is max-min expected utility, which recommends the policy for which the minimum expected utility, calculated by using the alternative probability distributions, is as large as possible. Another approach requires one to assign subjective weights to each of the alternative probability distributions and recommends the policy for which the weighted average expected utility is maximized. An advantage of these approaches is that they can incorporate multiple probability distributions that are consistent with available information without the need to select a unique probability distribution, as occurs with expected utility. These other approaches can also incorporate ambiguity aversion, for example, when a decision maker prefers a policy for which there is greater confidence about the probabilities. However, these approaches can be sensitive to the exact set of probability distributions that are considered, as well as to the assignment of weights to these distributions.

As discussed in Chapter 1, the committee interpreted its charge as focusing specifically on the SC-CO₂ for its use in regulatory impact analysis. The committee therefore developed its conclusions and recommendations to be consistent with an overall analytical approach based on the computation of expected net present value, taking into account that the SC-CO₂ is one of a large number of additional variables that enter into a typical regulatory impact analysis.

The committee recognized that the particular analytical framework chosen in developing the IWG SC-CO₂ estimates is based on probability-weighted present value. Although this framework is appropriate for its

application in regulatory impact analysis, it is not the only framework relevant to decision making under uncertainty in the context of national and international climate policy. The Intergovernmental Panel on Climate Change (Heal and Millner, 2014; Kunreuther et al., 2014) has described at length the advantages and disadvantages of alternative decision-making frameworks under uncertainty (including those cited above) for setting a carbon tax, determining a cap on carbon emissions, or employing other policy approaches.

The IWG’s purpose in calculating the SC-CO₂, however, is to provide estimates of the net damages from emitting 1 metric ton of CO₂. The SC-CO₂ estimates will be combined with other benefit and cost calculations for a regulation that affects CO₂ emissions—such as an energy efficiency standard for electric appliances. The uncertainties in the estimates of other regulatory benefits and costs with which the SC-CO₂ will be combined have been computed using the expected net benefit approach, which forms the basis for regulatory impact analysis under EO 12866 and OMB Circular A-4. For this reason, the committee also followed that approach.

Assigning Probabilities to Inputs and Outputs

Following the expected net benefit approach requires assigning probabilities to the outputs of each module.⁵ In general, the information exchanged between modules will be in the form of a distribution for each year (or other period) to facilitate Monte Carlo analysis: a frequency distribution, probability density function, or a set of values and associated probability weights that is representative of an underlying distribution. Chapter 3 outlines possible approaches to projecting future GDP, population, and emissions, using both the extrapolation of historical data and expert elicitation (see below). This approach would yield a set of GDP, population, and emissions projections that can be viewed as a representative sample from an underlying distribution. These values can then be used in the climate module, which generates, for each projection, a distribution of values of global mean temperature and other climate variables. For each socioeconomic projection and draw from the distribution of climate variables, the damages module can compute a distribution of damage estimates. Thus, the overall framework for SC-CO₂ estimation will have to be designed to support the large number of simulations that may be required for uncertainty analysis.

Uncertainty in the rate of per capita GDP growth can be reflected in the manner by which damages are converted to a present value in the dis-

__________________

⁵ For details on implementation, see Chapters 3-6.

counting module (see Chapter 6). In addition to this relationship between discounting uncertainty and uncertainty in observable variables (i.e., per capita GDP), discounting also often entails ethical judgments that are not reducible to a probability distribution. This additional variability in discounting is instead assessed through sensitivity analysis (see Chapter 6).

It would also be possible to use sensitivity analysis with respect to the probability distributions that are passed from one module to another, particularly those for which uncertainties are difficult to fully specify probabilistically. Chapter 3 describes an approach that could be used in the near term to derive a joint probability distribution over GDP, population, and emissions. One could use alternate probability distributions over these variables to explore the sensitivity of SC-CO₂ estimates to the probability distribution used.

EXPERT JUDGMENT IN SC-CO₂ ESTIMATION

Construction of any model requires some form of expert judgment to make choices among alternative functional forms, input variables, or other aspects of model structure that are consistent with available data and theoretical understanding. The effects of alternative choices on model results can be particularly important when extrapolating from the conditions under which a model is estimated or calibrated (which are necessarily conditions that have been observed) to the conditions relevant to estimating the SC-CO₂, which may be far in the future and involve climates, technologies, and other factors much different from those that have been observed.

“Expert elicitation” (or “structured expert elicitation”) is a method that can often prove useful in developing distributions over uncertain parameters or variables whose values need to be projected into the future. It is a formal process in which experts report their individual subjective probability distributions for an uncertain quantity. The committee believes that, for input variables having a limited empirical or theoretical basis for quantification of projections and their uncertainty, expert elicitation conducted according to best practices provides a useful and necessary approach. Expert elicitation is a method to characterize what is known about a quantity; it does not add new information as an experiment or measurement would. Ideally, it captures the best judgments of the people who have the most information and deepest understanding of the quantity of interest. For some quantities, there may be so little understanding of the factors that affect their magnitude that informed judgment is impossible or can produce only unreasonably wide bounds. Appendix C describes in detail methods for conducting expert elicitation.

PROCESS OF UPDATING THE ESTIMATES

Current U.S. government practice is vague regarding when and how a process of reviewing and updating the SC-CO₂ estimates might occur, which makes it difficult for stakeholders and researchers to anticipate future reviews and potential SC-CO₂ updates and to plan for the process. A regularized, institutionalized process would allow both groups to align their activities more sensibly. If the SC-CO₂ estimates are to reflect advances in scientific understanding of the climate impacts of greenhouse gas emissions and the economic impacts of climate change, a process is needed to assure that the SC-CO₂ estimates are updated on a regular basis. Regularizing the frequency of updates would help focus the attention of researchers on providing useful inputs to the SC-CO₂ process and would make the timing of updates predictable to agency staff and stakeholders.⁶

Because the SC-CO₂ is used in regulatory impact analyses for regulations that are being issued on a regular basis, the frequency of updates should balance the desire to incorporate improved scientific information with the need to allow for proper review of any changes. The frequency of updates needs to be short enough so that estimates of the SC-CO₂ do not lag too far behind the science while being long enough to allow significant new information to be generated and incorporated by the IWG and to allow for scientific peer review of the revised methods and estimates themselves. Moreover, the rate of scientific progress is variable and changes over time and is different for the many disciplines and fields involved. Overall, there is a need to balance the value of a regularized and predictable process with one that is rigidly prescribed.

__________________

⁶ Although the committee does not recommend that updates to the SC-CO₂ be tied to the release of assessments from the Intergovernmental Panel on Climate Change or the National Climate Assessment of the U.S. Global Change Research Program, it would be desirable for the IWG to take account of new evidence included in these assessments, as well as to communicate its information needs to those groups.

This process is illustrated in Figure 2-2. Step 1 involves the technical interagency process of updating SC-CO₂ estimates, taking into account recommendations for improvement from the scientific community and the public, scientific advances, as well as both internal government and external technical support and scientific peer review of individual modules to ensure that the proposed improvements accurately reflect evolving evidence and approaches. Incorporation of relevant external technical support and peer review of particular components (e.g., by experts in each of the module areas), and the overall framework and implementation, prior to public notice and comment would help ensure the scientific reliability and credibility of the estimates. The result of Step 1 would be a

draft update of SC-CO₂ methods and estimates. The committee estimated this step could take 2 to 3 years.

Step 2 involves obtaining input and comment on the draft update from both the public and the broadest possible scientific and technical communities, as well as other stakeholders. The result of Step 2 would be a finalized SC-CO₂ update for regulatory use that has incorporated public comments on the draft approach and estimates. The committee estimated this step could take 6 months to 1 year.

Step 3 involves a thorough independent scientific assessment of the SC-CO₂ estimation process, in order to track and assess new scientific literature over time and make recommendations for future improvements and research. The committee estimated this step could take 18 months to 2 years. The dotted box and lines at the center of the process indicate the multiple opportunities to incorporate research and scientific advances in the SC-CO₂ estimation process and for independent reviews to help inform research priorities.

The committee anticipates an overall process of roughly 5 years, which would allow 2-3 years between recommendations for improvements from an independent scientific assessment (end of Step 3) and the issuance of a draft SC-CO₂ update for public notice and comment (end of Step 1). Following from this recommendation, the committee has structured some of our other recommendations to distinguish those that we believe can be accomplished in the near term (2-3 years) from those that we believe will likely take longer to accomplish (i.e., “longer-term”). It is important that implementation of the research recommendations (in other chapters) proceed in parallel with the updating process described above.