2. Recommendation 1. Use both global and domestic measures of the social cost of greenhouse gases

Both global and domestic SC-GHGs are policy-relevant. Providing both estimates will increase transparency about the underlying analysis while also providing clarity about who receives the benefits, will foster policy discussions about fairness and equity, will furnish agencies with the flexibility to prepare analyses consistent with their statutory mandates, and will provide important distributional information to support international negotiations. We believe the emphasis given to the use of the domestic or global estimate depends on the context of the analysis and the basic principles of policymakers.

The exclusive focus on the global SC-GHG presumes that U.S. policymakers are indifferent to whether climate-control benefits occur in the USA versus elsewhere in the world – a perspective inconsistent with basic notions of national interest and the focus of U.S. laws, such as the Clean Air Act, that specify that relevant “benefits” are those to the “Nation” and not the world. It is also inconsistent with President Clinton’s 1993 Executive Order on regulation, which is still in effect, that seeks a regulatory system that serves the American people. Moreover, development and use of a domestic SC-GHG would increase transparency by providing information about reductions in domestic GHG-related damages from climate change initiatives to members of Congress, and to U.S. taxpayers and voters, who have a right to know the projected effect of policy decisions in the USA (Fraas et al. (Reference Fraas, Lutter, Dudley, Gayer, Graham, Shogren and Viscusi2016); Gayer and Viscusi, Reference Gayer and Viscusi2016).

Presenting a disaggregated picture of climate benefits is also consistent with commitments to consider the equity impacts of environmental policies. An exclusive focus on the global SC-GHGs is at odds with President Biden’s Modernizing Regulatory Review memorandum calling for more distributional analysis regarding “disadvantaged, vulnerable or marginalized communities” in the USA. The development of domestic SC-GHG estimates is a prerequisite for being able to monetize the impacts of GHG policies on such communities in the USA.

We agree with EPA that global SC-GHG estimates would help to promote future international cooperation and indicate the future scale of international cooperation that fighting climate change requires. But international cooperation and reciprocity must be appraised in terms of emissions reductions – that is, which countries can be expected to cut their emissions, by how much, and by what dates due to U.S. action – and an objective look at the likelihood of projected cuts. After all, most countries are failing to meet their commitments under the 2015 Paris Agreement, which did not establish legally binding emissions limits. Longstanding practice in regulatory analyses has been to incorporate only those changes in behavior that are required by current law. In the international context, a comparable approach would credit only reductions associated with binding agreements, not goals or pledges.

In its proposed revision of global SC-GHG estimates, EPA provides estimates for climate impacts physically occurring within the USA for a limited set of damage categories.Footnote ² These estimates are accompanied with the caveats that they only cover a subset of damages, do not capture spillovers or indirect effects, and are not equivalent to benefit estimates for US citizens and residents. These caveats are given as major reasons for presenting only global damage estimates.

These arguments are unpersuasive. Estimated damages at regional and country scales are aggregated into global estimates, so the global estimates have the same sectoral limitations as the damage estimates they are derived from. The EPA is confident enough in the aggregated estimates to equally weight them with the damage estimates derived from meta-analysis to provide a combined estimate for policy analysis (EPA-ERD, pp. 81). Moreover, country-level damage aggregates in the USA are discounted using a discount rate derived from U.S. market data, so there is a better match between the willingness to pay for climate damages and the discount rate used to convert them into present values in the USA than in other countries, where the U.S. rate is also applied. (discussed under Recommendation 4 below). Owing to this mismatch, SC-GHGs computed at the country level in the USA are likely to be more accurate than global damage estimates, given the sensitivity of SC-GHG estimates to the discount rate.

A rich set of economic and environmental data are available to support improved estimates of climate damages to the USA. As the science improves further, we would support developing and reporting damage estimates for states, regions, and counties in the United States. Modeling at regional scales could be “soft-linked” into the IAM – an increasingly common approach for combining complex models in the energy and climate areas (Krook-Riekkola et al., Reference Krook-Riekkola, Berg, Ahlgren and Söderholm2017).Footnote ³ A scientific advisory panel to the EPA recommended soft-linking as an option for incorporating disaggregated spatial environmental information into the EPA’s SAGE model (U.S. Environmental Protection Agency, 2017).

Looking to the long run, developing both domestic and global SC-GHG estimates is likely to promote stability in analytic practices at EPA and other agencies, reducing the likelihood of disruptive whipsawing of GHG policy analyses across administrations. Given the long-term nature of climate change policies, it is essential that the regulatory analyses provide data that will be of continuing policy relevance.

3. Recommendation 2. Use country-specific values for assessing nonmarket benefits

Analysts should evaluate the nonmarket benefits like mortality risk reductions using the willingness to pay of the populations in the affected countries. This makes the valuation of nonmarket benefits consistent with the valuation of the monetized portion of climate damages, which reflect country attributes as measured in GDP changes. The most widely developed estimates of international differences in the value of a statistical life (VSL) assess how the VSL varies with income. The EPA-ERD correctly recognizes that using the VSL in the USA for benefit transfer purposes without adjustment for income differences is not appropriate.

The value of mortality risk reductions is the largest benefit category in the SC-GHGs for the models that explicitly compute them.Footnote ⁴ The value of the mortality reduction benefits should reflect the willingness to pay of the residents in the countries where they occur. Preferences for risk reduction will differ across countries for a variety of reasons, such as social norms, religion, availability of health care, and income. The effect of income on the value of mortality risk reduction is the principal characteristic identified by the economics literature. As with other normal goods, the valuation of mortality risk reductions is an increasing function of income. If there is a positive income elasticity of the VSL, the mortality cost reductions in lower-income countries will be accorded a lower value.

The EPA uses an income elasticity of 1.0. The EPA cites a variety of references supporting a positive income elasticity. Some of these references are judgmental assessments or reports of recommended practices lacking empirical evidence. While some of the cited references explicitly advocate income elasticity values other than 1.0, a couple of the studies report empirical estimates of an international income elasticity of the VSL equal to 1.0.Footnote ⁵

In effect, the VSL accorded to mortality risk reductions in poorer countries will be a value that is proportionally lower than that used for the USA. Some observers have criticized income-adjusted VSL estimates for devaluing the lives of those outside the USA. This criticism is not warranted because a country’s income status influences its expenditure priorities. While there is disagreement over the recommended income elasticity, the EPA’s approach of applying a positive income elasticity to the benefits transfer of U.S. VSL estimates internationally is sound.

4. Recommendation 3. In the near term, adopt constant rates of 3% and 5% for computing SC-GHGs. Also use these rates to calibrate stochastic Ramsey discount rates

This recommendation gives four discount rate cases, ranging from lowest (stochastic Ramsey rate at 3%) to highest (constant discount rate at 5%). The constant rates are a return to the Obama-era rates. The 3% and 5% rates are quite conservative, given credit constraints in U.S. debt markets that many consumers use to smooth consumption. The stochastic Ramsey rate declines over time and is consistent with the point of view that declining discount rates are appropriate for intergenerational discounting.

4.1. Selecting target consumption rates of discount

Based on the evidence presented below, real consumption discount rates of 3% and 5% are quite conservative, given credit constraints in U.S. debt markets that many consumers use to smooth consumption, and are even less representative of consumption discount rates for low- and middle-income countries. However, to avoid an abrupt change in practice, we recommend that EPA use the 3% and 5% constant rates in the near term – a return to the practice of the Obama administration. Obama/IWG adopted 5% in part to provide that some “…account should be given to the discount rates revealed by their [consumer] behavior.” (IWG, 2010, p. 23).

The EPA-ERD selects a target consumption rate of discount of 2% bounded with a range from 1.5% to 2.5%. EPA justifies this range due to the substantial and persistent decline in the average real return to Treasuries in recent decades. However, real 30-year average Treasury rates were greater than 3% through 2013 (Figure 1). The 30-year average only dropped below 3% with the marked decline in rates after the Great Recession. Figure 2. presents 10-year Treasuries over the most recent 30-year period. The highlighted areas show periods when the Federal Reserve was lowering the federal funds rate -- the traditional Federal Reserve monetary policy to lower interest rates to address a weakening economy. After a recovery from the DotCom bust and 9/11, the real interest rates on 10-year Treasuries trended up from 2% to close to 3% before the Federal Reserve moved to address the Housing Market Crash beginning in September 2007 -- the crisis resulting in the Great Recession. The decline following 2007 is directly linked to the Federal Reserve’s “quantitative easing” policy – an unusual policy by historical standards – which operated to keep returns on Treasury notes artificially low (Kashyap and Stein, Reference Kashyap and Stein2023). In any event, interest rate behavior in the most recent three decades is not more relevant than interest rates from earlier historical periods, given the long-term forecasting horizon for SC-GHG estimates.

Figure 1. 30-Year Average 10-Year Real Treasury Rates.

Source: Preamble to March 2023 Proposed Revision to Circular A-4.

Figure 2. Measures of 10-year Real U.S. Treasury Rates.

Source: Preamble to March 2023 Revision to Circular A-4.

EPA also cites several studies that survey economists and other researchers in the climate area, that is, Drupp et al. (Reference Drupp, Freeman, Groom and Nesje2018), Howard and Sylvan (Reference Howard and Sylvan2020), and Pindyck (Reference Pindyck2019). However, the means and upper bound estimates of these surveys are generally higher than the target discount rates that EPA has proposed to use for its SC-GHG estimates.

The larger issue is that the market for 10-year Treasury notes is more relevant for infrastructure finance than for regulatory evaluation (Krutilla and Graham, Reference Krutilla and Graham2023). In contrast to public investments, the incidence of regulatory benefits and costs falls on some combination of companies that undertake capital investments for regulatory compliance, their customers, and the broader U.S. population. The market for 10-year Treasury notes is not an appropriate way to measure the discount rates of households whose consumption smoothing decisions are affected by climate regulations. Few individuals invest in 10-treasury notes (CEA, 2017); foreign governments, institutional investors, and the social security trust fund hold the largest share.

Individuals who do hold 10-year Treasuries are likely to be higher income. Data from the Survey of Consumer Finances show that very few low-income households hold securities with risk/ return/liquidity profiles like 10-year Treasury notes. In addition, they maintain only minimal balances in their transaction accounts – checking and savings accounts and the like – keeping only what is necessary for monthly expenses and small, unexpected expenses.

Data from the Federal Reserve’s most recent 2019 Survey of Consumer Finances indicate direct bond holdings are zero for more than 99 percent of families with incomes below the 60th percentile. Holdings of similar financial instruments are quite rare. Holdings of certificates of deposit are 2.7 percent, 5.7 percent, and 8.0 percent for the first, second, and third quintiles of family income. Holdings of savings bonds are 2.3 percent, 5.5 percent, and 7.7 percent for these three quintiles of families. Holdings of “pooled investment funds” are 2.1 percent, 3.6 percent, and 6.2 percent for these quintiles. Even retirement accounts, regardless of the type of financial assets included in them, are uncommonly held. For the first, second, and third quintiles of family income, 10.7 percent, 32.8 percent, and 53.7 percent of families have retirement accounts.

The EPA-ERD points out that credit-constrained consumers are likely to hold high interest rate debt and explains (pp. 53):

“This behavior may reflect rational intertemporal preferences, or it may reflect other factors such as present bias, lack of financial literacy, and other distortionary effects of poverty. …Nevertheless,… the high interest rates that credit-constrained individuals accept suggest that some account should be given to the discount rates revealed by their behavior.” EPA’s selection of target discount rates, however, does not give any weighting to revealed consumer behavior.

Field experiments have elicited consumer consumption discount rates directly. Nominal discount rates for U.S. military personnel making choices about compensation packages varied between 10% and 54% (Warner and Pleeter, Reference Warner and Pleeter2001). Other research reports implicit discount rates for used car buyers ranging from 2.8% to 16.9% (10th to 90th percentile) (Busse et al., Reference Busse, Knittel and Zettelmeyer2013) with a mean implicit discount rate of 13% (Allcott and Wozny, Reference Allcott and Wozny2014). Hausman reports a discount rate of 20% for consumers purchasing energy-using durables (Hausman, Reference Hausman1979).Footnote ⁶

Justifying regulations using social discount rates below consumers’ marginal rates of time preference has distributional implications (Miller, Reference Miller2015). Energy efficient appliances and vehicles that do not yield enough fuel bill savings to compensate for forgone current consumption, as consumers evaluate this tradeoff, burden consumers. Such regulations can incentivize consumers to delay new purchases, or to shop in secondary markets for used products at marked-up prices due to regulatory-induced demand shifts (Davis and Knittel, Reference Davis and Knittel2019). Raising the price of energy-derived products places singular burdens on low-income consumers. These effects go against the distributional equity objectives promoted in President Biden’s Modernizing Regulatory Review memorandum.

5. Recommendation 4. In the long-run, develop country-specific discount rates for intergenerational discounting

The assessed benefits of GHG policies in different countries should reflect the preferences in these countries, including rates used for intertemporal tradeoffs. Making the discount rates country-specific is consistent with the monetization of benefits, which are based on country-specific values. The vast majority of global consumers does not have access to the global capital market, and the local credit market conditions are likely to be as diverse as local climate effects. Country-level discount rates are more accurate than assuming the U.S. discount rate is representative of a global average. The consumption discount rates in low- and middle-income countries are likely to be significantly higher than those in the United States, where incomes are higher and credit constraints are relatively less binding. Policy analyses in some high-income countries use discount rates of 7–8%, and many low-income countries use discount rates in the 12% to 15% range (Campos et al., Reference Campos, Serebrisky and Suarez-Aleman2015).

EPA adopts a single global social discount rate for intergenerational discounting using only U.S. data. Yet, the vast majority of global consumers does not invest in 10-year Treasury notes, or have access to the global capital market. Many consumers in low-income countries transact in markets lacking formal financial institutions (Banerjee & Duflo, Reference Banerjee and Duflo2008). Even when financial institutions exist, the supply of credit can be limited, especially to low-income consumers, and lack of competition for financial services can lead to predatory lending practices. These factors raise consumer discount rates. Banerjee and Duflo (Reference Banerjee and Duflo2008) find that discount rates for both low- and middle-income consumers in less developed countries are about 4% per month.Footnote ¹⁰

Discount rates for lower income classes in low- and middle-income classes are often related to local conditions. For rural populations in Ethiopia, subjective discount rates rise/fall by 17% for a 10% loss/gain to income occasioned by droughts/high rainfall (Di Falco et al., Reference Di Falco, Berck, Bezabih and Köhlin2019). In other words, discount rates are endogenously related to the local effects of climate change. Mortality risks in six low-income countries raise discount rates. The median subjective discount rates vary from 38% to 45% on an annual basis over a five-year planning horizon (Poulos & Whittington, Reference Poulos and Whittington2000).

Country-specific discount rates have been estimated in Europe for benefit–cost analysis of projects financed from structural adjustment funds (Florio & Sartori, Reference Florio and Sartori2013). As a longer-term project, consumption discount rates should be estimated in every country with damage estimates that enter global SC-GHG estimates. Country-level discount rates should be used with country-level growth rates and damage estimates to estimate country-specific SG-GHGs. This would make the discounting part of the economic evaluation internally consistent with the use of country-specific values for generating and valuing physical damage estimates. Undertaking this task may seem daunting but should be considerably less difficult than estimating local and country-level damages, as is done using the DSCIM and GIVE damage functions. As explained in the introduction, the pay-off of improving discount rate estimates is high, given their impact on the magnitude of SC-GHG estimates.

6. Recommendation 5. For discounting in climate rules, adopt a two-step procedure that first annualizes fixed investment costs using the social cost of capital, and then discounts benefits, operating costs, and the annualized fixed cost streams using the consumer’s rate of time preference

The effects of climate regulations on both consumption and capital formation need to be addressed consistent with using a shadow price of capital. For the short term, the recommended SOC rate is a 7% real rate of interest. In the long run, sector-specific rates of return on capital should be developed, given that risks and returns differ among different sectors impacted by regulation. For the second stage, CDRs of 3% and 5% are recommended as conservative estimates for consumer discount rates.

OMB’s Circular A-4 states that “the analytically preferred method of handling temporal differences between benefits and costs is to adjust all the benefits and costs to reflect their value in equivalent units of consumption and to discount them at the rate consumers and savers would normally use in discounting future consumption benefits” (OMB, 2003). The “shadow price of capital” is used to convert investment costs into consumption equivalents. Appendix A2 in the EPA-ERD “Consumption Rate of Interest and Integration into Benefit–Cost Analysis” develops a formula for the shadow price of capital that is consistent with the OMB’s recommended methodology (EPA-ERD, pp. 112–115).

Kolb and Scheraga (Reference Kolb and Scheraga1990) show that annualizing investment costs using the SOC, and then discounting benefits, operational costs, and the annualized capital costs using a CDR is consistent with using a shadow price of capital. This method is understandable and tractable to implement.

The Kolb/Sheraga method is particularly apt for the analysis of environmental regulations. It assumes that investment costs are exclusively in the form of displaced private capital, which fits the common scenario that companies incur capital investments to meet regulatory requirements. Companies also incur operational costs for running capital equipment, or from process adjustments in response to regulatory requirements, for example, boiler maintenance, fuel-switching, and the like. The Kolb/Sheraga method assumes operational costs are passed onto consumers, which is a reasonable assumption.

We encourage the EPA to follow this tractable approach. In the short term, 7% would be a reasonable SOC for annualizing capital costs. Estimates over the last 50 years for the annual rate of return to capital have been remarkably stable, with the rate centering around 7 percent (CEA, 2017, p. 10). In the longer term, sector-specific rates should be developed – a practice used by the French government. These rates will reflect risk premia (Damodaran, Reference Damodaran2022). Sector-specific rates are likely to fall in the range of 5 to 10%. As explained above, conservative CDRs to implement the second stage would be 3% and 5%.

7. Recommendation 6. Provide probabilistic uncertainty analysis

Uncertainties around input parameters should be traced through to SC-GHG estimates. Distributions for the SC-GHG estimates should be reported for use in RIAs.

The EPA-ERD represents uncertainties as distributions for uncertain input parameters like population size, economic growth, discount rates, surface temperatures, and the like. However, it does not report how the uncertainties associated with the input distributions trace through to distributions of the SC-GHG estimates themselves, except for one illustrative year – 2030. Given the input uncertainties documented in the EPA-ERD, SC-GHGs should not be presented as central values for the years beyond 2030.

There are several options for better representing uncertainties in the estimates of GHGs. The estimates could be presented as distributions for use in Monte Carlo simulation. Combined with distributions for other uncertain parameters, like concentration-response coefficients for PM2.5, this would make Monte Carlo standard in RIA.Footnote ¹¹ Alternatively, quintiles or quartiles of the SC-GHG estimates from an integrated distribution could be developed. In all cases, presentation of distributions should include lower bound values as well as upper bound estimates.

8. Recommendation 7. Conduct a temporal distribution analysis

This distributional assessment should separately report the fraction of the SC-GHG value from the present to 2100; from 2101 to 2200; and from 2201 to 2300. The uncertainty range for SC-GHG estimates should be displayed in each of these time intervals.

Presenting a separate assessment of SC-GHG values occurring at different points in the future would increase transparency, indicating how uncertainties are distributed over different time periods. This distributional representation would reveal the importance of benefits farther in the future relative to those closer to the present, allowing the public to see how the benefits are distributed among different generations. Knowledge of the benefits of GHG emissions reductions falling on the current generations is policy-relevant. The public bearing the costs of regulations is entitled to know the benefits of GHG reductions realized by the current generation and by future generations.

9. Concluding remark

Estimating the economic cost of global damages requires judgments about valuation, the representation of intergenerational tradeoffs, and the appropriate reporting format for policy analysis. This article adopts the view that transparency about normative perspectives, rationales, and uncertainties associated with SC-GHG estimates is in the public interest. As such, we have developed seven recommendations that will better inform the public and decision-makers about the value of reducing climate risks. These include recommendations to report both country-level and global damage estimates, to use country-specific values for monetizing climate damages, to represent SC-GHGs as distributions rather than as central values, and to conduct a temporal distributional analysis that shows how climate benefits are distributed among different generations. We also provide a suite of recommendations for appropriate discount rates for estimating SC-GHGs, and for the RIAs that incorporate them. Discount rates have a singular impact on SC-GHG estimates, given that benefits occur over several centuries. Hence, the choice of an intergenerational discount rate is a crucial normative judgment in the analysis. The discount rates proposed in our recommendations are based on consumer discount rates in the credit markets that are most relevant for RIA.