2.1 Introduction

A regulation places legal obligations or constraints on those covered, usually individuals, businesses, or other organizations in society. Regulation is different from a public investment program, where the government raises funds and uses those funds to pay the costs of the program. For a regulation, the costs of compliance are typically incurred – at least initially – by the individuals and/or organizations subject to the regulation. Over time, regulatory costs to businesses may be passed on to consumers in the form of higher prices for products and services, or they may be paid for by diminished compensation to employees or owners of businesses.

Some regulations have strong policy rationales (e.g., civil rights protection), but the rapid proliferation of regulations in the twentieth century led to increasing concerns among businesses and the public. Critics argued that regulation was hurting the performance of the economy by contributing to inflation, reducing productivity, and curbing market innovation. Advocates of “regulatory reform” sought to eliminate unnecessary or overly burdensome regulations. When a regulation is justified, reformers promoted “smart regulation” that accomplished societal goals in ways that minimize cost and preserve as much flexibility as possible for personal and business choice.

One outgrowth of the reform movement is the requirement that regulators commission RIAs, a key element of evidence-based approaches to policymaking (Reference Radaelli, Harris, Bitonti, Fleisher and BinderkrantzRadaelli, 2020). An RIA may be undertaken before a new regulation is imposed and/or after a regulation takes effect, to ensure that the intended outcomes are achieved at reasonable cost.

The OECD defines RIA as a critical assessment of the positive and negative effects of proposed and existing regulations and nonregulatory alternatives (OECD, 2021). This definition is broader than BCA, even though BCA is a commonly used tool in RIA. Other methods include cost-effectiveness analysis, formal uncertainty analysis, scenario analysis, multiobjective decision analysis, risk assessment, and distributional analysis.

Some form of RIA is now commonplace in all OECD countries, but the practice of RIA is relatively recent in many developing countries (OECD, 2021).

2.2 Evolution of Regulatory Impact Assessment in the United States

President Ronald Reagan’s 1981 Executive Order 12,291 is often cited as the origins of RIA in the United States even though presidents Gerald Ford and Jimmy Carter were both strong proponents of rigorous regulatory analysis. Executive Order 12,291 required federal agencies to prepare RIAs in support of proposed and final regulations. It also prohibited agencies from publishing proposed and final rules in the Federal Register without clearance from the Office of Management and Budget (OMB), which enabled OMB analysts to negotiate changes to rulemaking packages, including RIAs (Reference GrayGray, 1998; Reference Miller IIIMiller, 2011).

The Reagan administration’s implementation of Executive Order 12,291 triggered substantial controversy and negative press for the Reagan White House. Critics charged that Reagan was more interested in a “relief” program for business than a constructive reform of regulatory process and policy (Reference Eads and FixEads and Fix, 1984; Reference OlsonOlson, 1984). Some Reagan appointees at the agencies were slow to respond to legislative deadlines for new regulations.

The Reagan administration, facing a possible cutoff of funding for the Office of Information and Regulatory Affairs (OIRA), compromised by making the OIRA review process more open to public scrutiny and by agreeing to subject future OIRA administrators to a formal Senate confirmation process. Under president George H. W. Bush, the Senate did not confirm Bush’s OIRA nominee, so OIRA staff worked informally with a special White House office, the Council on Competitiveness, overseen by vice president Dan Quayle. Agencies such as the Occupational Safety and Health Administration and Environmental Protection Agency (EPA) persistently resisted the efforts of OIRA staff to make changes to their RIAs and rulemaking documents. Gradually, as the unitary theory of executive power became better developed, White House regulatory review through OIRA became better accepted (Reference SunsteinSunstein, 2012).

Under presidents Bill Clinton, George W. Bush, Donald Trump, and Joe Biden, implementation of the RIA aspects of the federal rulemaking process stabilized with some important exceptions. A Clinton executive order narrowed the OIRA review scope to “significant” regulations but granted OMB broad authority to decide which rules are significant. Clinton called for more consideration of distributional equity issues in RIA and replaced Reagan’s numerical benefit-cost test with a more nuanced “benefits justify costs” test. The word “justify” is seen as allowing agencies to consider qualitative benefits and costs as well as equity issues (Reference KatzenKatzen, 2018). Under George W. Bush, OIRA largely retained the Clinton approach but issued a technical guidance document on how to perform RIA, guidance that remains in effect today (OMB, 2003). The Obama administration emphasized the review of existing regulations and incorporated more behavioral economics into RIAs (Reference SunsteinSunstein, 2014). The Trump administration launched an ambitious deregulatory agenda but much of this agenda was lost in court, in part due to poor-quality RIAs that did not adequately consider the foregone benefits of regulation (Reference Belton and GrahamBelton and Graham, 2019). The Biden administration repealed the Trump administration’s “regulatory budget” (aimed at spurring deregulation) but also announced it will retain the RIA process overseen by OMB. It pledged to modernize OMB Circular A-4, in part to give more emphasis to justice/equity concerns.

2.3 Evolution of Regulatory Impact Assessment in Europe

The United Kingdom and the Netherlands pioneered the “Better Regulation” agenda in Europe, leading to the European Union’s eventual embrace of the agenda. The term “Better Regulation” has no universal definition but typically encompasses ex ante RIA, transparency in regulatory development, reduction of the administrative costs of regulation, consultation with stakeholders, and proportionality in the regulatory response to a problem (OECD, 2019; Reference WienerWiener, 2006).

In 1985, the UK Government initiated a requirement that regulators prepare compliance cost assessments, emphasizing impacts on businesses and the economy. A broader RIA requirement, including benefits and costs, was adopted in 1998 and remains in effect today. A full RIA must accompany any proposals for primary or secondary legislation when they are proposed to the Parliament.

The Netherlands initiated business-impact studies in the 1980s. In 1994, the Dutch government called on regulators to streamline regulation “to what is strictly necessary.” In the late 1990s, analysts in the Dutch government developed a new “Standard Cost Methodology” for “administrative costs” – what in the United States politicians call “red tape” or “paperwork burden.” Using this metric, the Dutch government has accomplished one round of 25 percent reduction in administrative costs. However, administrative costs typically account for a small share of the overall business or societal costs of regulation, and the Dutch government is moving toward more comprehensive ex ante RIAs that quantify benefits and costs of regulatory proposals.

The European Union was slow to embrace Better Regulation until the Lisbon Agenda’s emphasis on a competitive European economy in March 2000. Several years later, under the leadership of the Italian politician and economist Romano Prodi, the European Commission issued preliminary guidance on how to conduct RIAs. In 2005, under the leadership of Portugal’s former prime minister, Jose Manual Barasso, the Commission instituted a mandatory RIA requirement and empowered a new Impact Assessment Board to review the quality of Commission RIAs, before they were released to the Parliament and the Council. With minor modifications, the EU process established by Barasso remains in effect today. In general, the EU embrace of RIA entailed less controversy than occurred in the United States, in part because the EU was able to learn from and respond to the US experiences (Reference GolbergGolberg, 2018)

2.4 Comparing Regulatory Impact Assessment Requirements in Europe and the United States

The scope of the EU’s RIA requirement is much broader than the scope in the United States (Reference Wiener and AlemannoWiener and Alemanno, 2010). The US RIA requirement does not apply to primary legislation considered by Congress; it applies only to the implementing regulations developed by the executive branch. In Europe, RIAs are required for all proposals of primary legislation as well as secondary legislation, regulations, directives, communications and White Papers.

On the other hand, the results of RIAs have more potential impact on decision-making in the United States than they do in the EU. The United States uses OMB as a centralized authority to ensure that regulators perform RIAs and use the results of RIAs in rulemaking. The OMB is involved in regulatory policymaking as well as analysis quality. The EU’s centralized process focuses primarily on analysis quality, delving less frequently into the substance of policy making choices. As an agent of the president of the United States, OMB has more authority than the EU’s centralized review body – now called the Regulatory Scrutiny Board (formerly the Impact Assessment Board). As a nationally elected politician, the president of the United States has more power than the president of the Commission, who is selected by the heads of state in Europe, and this difference reveals itself in the usually strong powers of OMB relative to US regulatory agencies (Reference KaganKagan, 2001).

Judicial review of RIAs, and their use in regulatory decisions, is also a bigger factor in the United States than in Europe (Reference Bull and ElligBull and Ellig, 2017). Regulatory Impact Assessments in the EU rarely play a role in litigation after a regulation is adopted but in the US RIAs are often considered in judicial review of a final regulation (Reference Cecot and ViscusiCecot and Viscusi, 2014). Through the Administrative Procedure Act, the US Congress has empowered any individual or organizations harmed by a new regulation to challenge its legitimacy in federal court, and US courts tend to encourage benefit-cost reasoning (Reference Noe and GrahamNoe and Graham, 2019). In Europe, absent compelling circumstances, judicial review of a new regulation is available only to one or more of the twenty-eight member states of the EU (Reference Stone SweetSweet, 2003).

The technical quality of RIA is highly uneven in both the United States and Europe (Reference Dudley and MannixDudley and Mannix, 2018; Reference Fraas and LutterFraas and Lutter, 2011; Reference Radaelli and De FrancescoRadaelli and De Francesco, 2013). The quality challenge has magnified as many of the twenty-eight member states in the EU and the fifty states of the United States have initiated RIA programs of various degrees of comprehensiveness and technical sophistication.

In the following sections, we shift from the institutional setting of RIA to the methodology issues associated with formulating regulatory BCA, particularly, BCA applied to air pollution, energy, and climate change regulations. In the United States, the benefits of AEC regulations have dominated the total benefits of the US federal regulatory program, and BCA of these regulations has generated significant political controversy and associated legal actions from environmental and industry groups. The role of BCA in the evaluation of AEC policies and regulations in Europe have been less significant than in the United States, but the use of BCA continues to advance with EU initiatives like the EU’s Clean Air Policy Package and country initiatives to reduce air pollution and implement commitments under the Paris Accord.

3.1 Introduction

Regulatory compliance diverts resources from other activities in the economy, imposing opportunity costs. Costs occur in the present, and also in the future when a regulation affects savings, investment, and capital accumulation. Forecasts of future regulatory impacts can also affect current-period economic adjustments.

The distributional pattern of regulatory costs affects their estimation. As noted in Section 2, regulatory costs are “off budget,” falling on multiple actors in the economy. The incidence of compliance costs can be shifted, as when a polluter passes on some costs to consumers; regardless, costs are ultimately borne by the private sector without compensation.

Different kinds of regulatory costs can be identified. Direct “abatement costs” are the amount by which the costs of output increase over a given range. Partial equilibrium (PE) costs add the opportunity costs associated with output and price adjustments in the regulated market. General equilibrium (GE) costs incorporate the adjustments and feedbacks among all markets in the economy impacted by regulatory compliance.

For both practical and conceptual reasons, the scope and specificity of regulations influences the costing method. Some regulations target specific products or processes, such as appliance efficiency standards, rules governing the accidental releases of methane from natural gas pipelines, or pollution control requirements for refinery emissions. At the other end of the spectrum, regulations can apply economy-wide, such as a permit trading system used to implement a country’s carbon emissions target.

For differentiated regulations of limited scope, engineering cost assessments of abatement costs is the feasible method, given data availability and limited budgets for conducting economic analyses. The proportionality requirement for RIA in the EU is less stringent than in the United States, and engineering cost estimates are likely to be used for many regulatory analyses in the EU.

Partial equilibrium models are useful when a fine-grained resolution of the regulated market is necessary for assessing the structure of regulatory alternatives and their market impacts (e.g., see Reference AbitoAbito, 2020). Partial equilibrium modeling can also assess the dynamics of market behavior, such as regulatory effects on savings and investment, firm exit and entry, and industry concentration (e.g., Reference Fowlie, Reguant and RyanFowlie et al., 2016).

General equilibrium modeling is conceptually justified when the regulated market is linked to other markets, and these markets are distorted (SAB, 2017). Linkages result when the good produced in the regulated market is a substitute or complement to goods produced in other markets, and/or the regulation affects equilibrium conditions in input or output markets upstream or downstream in the supply chain. The welfare cost of distortions in these markets will increase or decrease as their equilibria changes, depending on the nature of the distortion and the direction of the market adjustment (Reference HarbergerHarberger, 1964).

From a conceptual point of view, linkages to distorted markets, rather than the size of the regulation, determine the relevance of GE approaches (SAB, 2017).Footnote ² From a practical point of view, data constraints limit the specificity with which policy options and the structure of regulated markets are represented in current-generation computational general equilibrium (CGE) models, making them difficult to apply in the RIA of many kinds of regulations. The continual evolution of methods is likely to make CGE modeling more routine in RIA, building from the large academic literature that shows the significance of “second-best” welfare costs for regulatory evaluation.

We review each of these costing approaches in the subsections that follow. Costs are taken as the measurement of changes between static equilibria in the presence and absence of the regulation, or in dynamic models, between steady states or balanced growth paths. Factors are assumed to be fully employed before and after the regulation (this assumption is reconsidered in Section 6 on distributional effects). Structural rigidities in the economy that reduce the mobility of inputs, such as sector-specific capital or labor, can be reflected in this assessment, but transactions costs arising from property rights exchanges are excluded (these costs are also addressed in Section 6). The economic implications of transition paths between steady states or balanced growth paths in dynamic models are not assessed in the literature reviewed in this section.Footnote ³

3.2 Estimating Abatement Costs

Abatement costs are derived from “engineering cost” estimates that compile the cost of each of the resources that are estimated to be used in compliance, both one time and recurring. Market prices, which are often available from surveys of polluters and/or vendors of pollution control equipment, are used to represent unit values.Footnote ⁴ Vendor prices sometimes differ from those stated by polluters, but a range of estimates can be used. In complex cases, a regulatory agency can contract with a consulting firm to provide engineering cost estimates of pollution control technologies.

Direct abatement costs are affected by the type of regulation. Technology-based standards or stringent performance standards compel particular “end of pipe” control methods such as catalytic converters or particle traps to reduce vehicular emissions, or wet and dry scrubber technologies to reduce point-source emissions. Even with the specification of technologies, however, there may be more than one compliance option. The engineering cost algorithms used in EPA’s greenhouse-control RIAs for vehicular emissions contain literally hundreds of control technologies that can be combined in various ways depending on the market segment. Forecasting the cost-effective technology choices for heterogenous polluters facing a variety of technological compliance options can be difficult.

When regulations cause a technology to be used on a larger scale than the historical pattern, the supply chain of the technology may experience cost savings due to learning by doing and economies of scale. When forecasting the unit costs of mass-produced technology, the US EPA often incorporates gradual rates of savings to account for these processes.

It is not uncommon for emissions regulations to stimulate technological innovation in pollution control and / or monitoring technologies, lowering the cost of regulation. The EPA’s original tailpipe standards in the 1970s induced costs due to new designs of motor vehicle engines but the performance advantages of the new engines produced nonpriced advantages that, when valued using hedonic methods, were larger than the market costs of the engines.

Experience with new technologies is not always better than anticipated. A new technology may not perform as well as vendors projected or may produce side effects that were not anticipated (e.g., sulfuric acid emissions from the early catalytic converters). Some fuel-efficiency technologies alter the way a vehicle feels when driven, and consumers have rejected some of these technologies.

Performance standards are another approach to controlling pollution. These standards impose emissions limitations without specifying the method to achieve them, offering more compliance flexibility than technology-based standards. A performance standard can limit total emissions, or the level of pollution produced per unit of output or input. For example, standards can limit the usage of energy per megawatt hour of electricity or limit the mass of pollutants emitted per unit of energy input. These differences lead to different economic adjustments, giving different abatement costs (Reference HelfandHelfand, 1991).

An important justification for performance standards is asymmetric information. Polluters know more about their abatement options than regulators, giving polluters the flexibility to exploit their private information to choose the combination of pollution control methods that minimize their costs. To comply with an emissions standard, for example, a stationary-source polluter might increase boiler maintenance, change boiler running times or operational conditions (e.g., for NO_x emissions control), or fuel switch, altering the mix of operational inputs per unit of output. Pollution controls are another option. These cost-minimizing choices can be difficult to forecast.Footnote ⁵ As with more prescriptive approaches, it is not uncommon for analysts to make educated guesses about polluters’ compliance strategies.

Incentive-based instruments like emissions taxes or tradable permits allow the additional flexibility for cost-minimization across a population of polluters, broadening compliance options to include differential emissions control among them. Regulations that specify allowable air pollution concentrations over a geographic area, usually averaged over a period of time (e.g., twenty-four hours to one year) give the most compliance flexibility. National Ambient Air Quality Standards in the United States offer an example of this type of regulation. To comply with ambient standards, abatement tradeoffs can be made among different sectors and end uses.

If priced-based policy instruments like emissions taxes or tradable permits are implemented, the marginal costs of pollution control are revealed in the market price of pollution. This property avoids the need to estimate abatement costs, an important advantage in addition to the cost-efficiency of these approaches. On the other hand, forecasting compliance modalities for ambient air quality standards is a significant challenge. As a result, abatement cost estimates for ambient standards come with a large uncertainty bound.

3.3 Behavioral Effects and Partial Equilibrium Costs

Direct abatement cost estimates do not reflect market responses to regulation. In the general case, the equilibrium output level in regulated markets will decline in response to regulatory actions. This “output substitution effect” provides an additional margin for cost-minimization, lowing the social cost of the regulation (Reference Marten, Garbaccio and WolvertonMarten et al., 2019; Reference Pizer and KoppPizer and Kopp, 2005). Total costs now include abatement costs on inframarginal output produced after the regulatory adjustment, and the net of producer cost savings and consumer losses over the range that output declines.

The output substitution effect depends on the degree to which the regulation increases direct abatement costs and on how the market responds. GE feedbacks affect both. In general, cost increases reflect the type of regulation and abatement options, as previously discussed, and characteristics of production processes, such as the structure of the production function, the type of inputs and their share in production, and input substitution possibilities.Footnote ⁶ The effect of higher costs on the output level is determined by GE supply and demand relationships when the regulated market is competitive. If the market is not competitive, the strategic behavior of polluters also becomes a variable affecting the output response. When the regulated product is traded, the world excess demand for output (which can be positive or negative) also influences market behavior (Reference KrutillaKrutilla, 1991).

The impact of regulations on market prices – regulatory cost “pass through” – under competitive and noncompetitive market structures is the subject of a large literature that studies energy price signals in response to carbon emissions restrictions; the fraction of carbon rents required to compensate the regulatory costs borne by polluters; and the effects of pricing and rent-sharing strategies on the competitiveness of regulated sectors in the international market (Reference Fabra and ReguantFabra and Reguant, 2014; Reference Neuhoff and RitzNeuhoff and Ritz, 2019; Reference Sijm, Chen and HobbsSijm et al., 2012). Whether the market predictions in regulatory cost models – either PE or GE – are consistent with the findings of this literature could provide a useful check on modeling assumptions.

In OECD countries, concentrated industries or regulated public utilities account for a significant share of energy use, local air pollutants, and CO₂ emissions. PE models can model the structure of such markets to capture the impact of strategic producer behavior. An example is a study of policies to reduce CO₂ emissions from the cement industry in the United States (Reference Fowlie, Reguant and RyanFowlie et al., 2016). This industry is an oligopoly that faces a competitive fringe of foreign suppliers.

The study uses a dynamic infinite horizon model in which market participants compete over quantities in the present period and invest (or disinvest) in future capacity. In the present period, emissions restrictions reduce domestic output, worsening the terms of trade and the oligopolistic market distortion. In the longer term, reduced profits induce firm exit from the industry, increasing market concentration. In this setting, auctioned permits are shown to impose relatively high welfare costs. Grandfathered permits impose the same short-run efficiency costs, but the prospective value loss of foregone free permits reduces the incentives for firms to exit. Pairing a border tax with a permit auction attenuates negative terms-of-trade effects while providing tariff revenue. This lowers welfare costs in the short and long run relative to permit auctioning alone. A dynamic updating scheme – the equivalent of granting permits in proportion to the firm’s previous period’s output – confers an output subsidy that reduces the oligopolistic market distortion and the terms-of-trade deterioration. As is the case with grandfathered permits, dynamic updating also reduces the incentive for firm exit. In all, market structure, trade adjustments, and intertemporal dynamics are shown to influence the relative costs of alternative policy designs.

3.4 Distorted Secondary Markets and General Equilibrium Modeling

Starting with research on environmental tax reforms in the mid-1990s (e.g., Reference Bovenberg and De MooijBovenberg and de Mooij (1994), Reference Bovenberg and GoulderBovenberg and Goulder (1997), Reference BovenbergBovenberg (1999)), a large literature has documented the second-best welfare effects of emissions restrictions (e.g., Reference Goulder, Parry, Williams III and BurtrawGoulder et al. (1999), Reference Parry and Williams IIIParry and Williams (1999), Reference Parry, Williams III and GoulderParry et al. (1999)). This research suggests that CGE modeling is necessary to trace through the linkages in the economy that affect costs when regulations impact distorted markets. The GE framework also produces conceptually correct welfare estimates, and provides an internally consistent representation of the economy, in the sense that all markets clear at all times, and all agents meet their budget constraints (SAB, 2017).

Constructing CGE models involves time and technical expertise, however, and data constraints often limit the granularity of the economic representation. To make the modeling tractable, CGE models typically aggregate consumers, production sectors, and geographic regions to a manageable number (Reference Carbone, Bui, Fullerton, Paltsev and WingCarbone et al., 2022). Aggregation as such does not significantly affect welfare cost estimates averaged over the entire economy (Reference Fullerton and TaFullerton and Ta, 2019). However, to be useful for regulatory cost estimation, models must be finely enough revolved to represent the processes or products that regulations target, and to describe the regions in which regulations are implemented if regulations differentially target specific areas, for example, those with the highest local ambient air pollution concentrations. Models should also be able to capture abatement cost heterogeneities that affect regulatory costs.

An option to address these issues is to selectively aggregate portions of the CGE model that do not need to be represented in detail, while highly disaggregating the sectors of relevance (Reference Carbone, Bui, Fullerton, Paltsev and WingCarbone et al., 2022). Another strategy is to incorporate highly disaggregated bottom-up or PE models within a CGE model. Methodology research on model linking is rapidly growing in the energy and climate area. So-called “soft-linking” combines models by running each of them separately, with outputs of one model serving as inputs into another. The models are run iteratively until a solution is reached. Soft-linking is a tractable way to integrate two models that are computationally burdensome even when solved alone (Reference Krook-Riekkola, Berg, Ahlgren and SöderholmKrook-Riekkola et al., 2017). However, the compatibility of modeling structure is an issue when soft-linking off-the-shelf models. An alternative is to construct the bottom-up and CGE models from scratch (Reference Andersen, Termansen, Gargiulo and GallachóircAndersen et al., 2019), or to “hard-link” models by integrating them completely (e.g., Reference Helgesen, Lind, Ivanova and TomasgardHelgesen et al., 2018). Hard-linking imposes computational burdens that may be insuperable. With continuing research, however, methodology advances are likely.Footnote ⁷

A regulatory agency could develop a suite of models that include PE and GE models that aggregate sectors in different ways for different categories of regulatory analyses (SAB, 2017). Using different models to provide cost estimates for the same regulation would also increase information for RIA.Footnote ⁸

We now turn to academic research that illustrates the insights from the GE literature on regulatory cost estimation in distorted economic settings. The first topic is second-best effect of existing taxes in capital and labor markets (the so-called tax interaction effect). This subject has received the greatest share of research. The second topic is the second-best effects associated with noncompetitive markets.

4.1 Introduction

In this section, we focus on the benefit side in RIA. The standard approach is to model the link between regulatory interventions and health or environmental outcomes and then to apply unit values to the outcomes, giving total benefits. Market prices, market-derived shadow prices, or benefit transfers, such as the value of statistical life (VSL) or the social cost of carbon (SCC), are typically used to represent unit values.

This standard approach estimates the benefits outside the economic model that generates the costs. This dichotomy has conceptual limitations. Including environmental goods with other goods nonseparably in utility in a CGE model would give a theoretically correct demand system for all goods, reflecting GE feedbacks and the economy’s budget constraints (SAB, 2017). Benefit valuations derived within such a model are likely to differ from those conventionally estimated. For example, the standard approach relying on constant unit values implies nondeclining marginal valuations, which is inconsistent with demand theory. Conventional approaches also pose the risk of double counting, if the scope of different benefit categories is not clearly delineated.

It may be more feasible to specify the impacts of morbidity in CGE models than preferences for mortality risk reduction, see for example, Reference Mayeres and Van RegemorterMayeres and Van Regemorter (2008). As a longer-term goal, a scientific advisory board of the US EPA recommends representing environmental goods nonseparably in utility and calibrating preferences for mortality risk aversion in EPA’s numerical GE models (SAB, 2017). To be useful for valuing the benefits of many types of regulations, CGE models would also need to be disaggregated, or linked to other models, to represent policy-relevant spatial heterogeneities, such as local exposure risks and demographic characteristics.

In the remainder of this section, we stick with the conventional practice and, given space constraints, limit the scope of this review. In the next subsection, we overview the pathways that give rise to the benefits of AEC regulations. We then turn to two benefit categories that have received significant attention from researchers and policymakers: the value of reducing risks of illness and premature death, and the value of reducing greenhouse gas emissions. Primary valuation methods are not common for regulatory evaluation and are not considered here.Footnote ¹⁷

4.2 Benefit Valuation Pathways

Figure 1 shows the various pathways for benefits arising from AEC regulations. Energy efficiency regulations or climate policies like carbon emissions standards reduce greenhouse gases (arrow 1). Air pollution regulations that reduce local or regional exposures to black carbon also reduce the risks of climate change (arrow 2). The value of carbon emissions reductions are monetized using the SCC or target-consistent carbon prices (TCPs). Other greenhouse gases typically are monetized from their carbon equivalent (discussed in Section 4.4).Footnote ¹⁸

Energy savings are another benefit category from energy efficiency or climate regulations (arrow 3). In most RIAs, market prices are used to value these benefits, sometimes with adjustments for “rebound effects” and/or market distortions (HM Reference TreasuryTreasury, 2021; Reference Krutilla and GrahamKrutilla and Graham, 2012).Footnote ¹⁹

Figure 1

Benefit valuation pathways

Air pollution is reduced when fossil-fuel combustion declines or regulations target emissions directly. Thus, energy and climate regulations can reduce local or regional air pollution as a co-benefit (arrow 4), while reductions in local air emissions are the primary effect of air pollution regulations (arrow 5). A Chemical Transport Model (CTM) is commonly used to map air emissions reductions at local sources into pollutant concentrations at local receptors (arrow 6). Concentration- or exposure-response functions are then used to translate changes in air pollutant concentrations or exposures into health or environmental effects (arrows 6/7-8; 6/7-9; 6/7-10).

The left-most branch for “Exposure Reduction” indicates the impact of local air quality on agriculture and other renewable resources (arrow 8), for example, ground-level ozone impacts agricultural productivity (Reference Hong, Mueller and BurneyHong et al., 2020; Reference Sun, Yun and YuSun et al., 2017). Air pollution also damages materials, buildings, and historical sites (Reference RablRabl, 1999; Reference SpezzanoSpezzano, 2021) (arrow 9). Market prices or market price-derived shadow prices are usually used to monetize the benefits of reducing these impacts.

Exposure reductions also provide a variety of benefits associated with improving human health (arrow 10). Health care resource savings are generally monetized with market prices, or in some cases shadow prices that remove price markups (Reference Drummond, Sculpher, Claxton, Stoddart and TorranceDrummond et al., 2015). Health improvements also increase labor productivity and affect utility directly (Reference Mayeres and Van RegemorterMayeres and Van Regemorter, 2008). Wage rates are typically used to value productivity gains. In applied work, the value of leisure is sometimes monetized as fraction of wage rates. But because time savings are important in wide variety of economic contexts (e.g., transportation and recreation) there is large literature on estimating the value time, for example, Reference Dalenberg, Fitzgerald, Schuck and WicksDalenberg et al. (2004), Reference Jara-Díaz, Munizaga, Greeven, Guerra and AxhausenJara-Diaz et al. (2008).

Reducing air pollution exposures reduces a variety of morbidity endpoints including chronic bronchitis and nonfatal heart attacks. These can be monetized using market prices, market price-derived shadow prices, or using stated preference methods, for example, Reference Viscusi, Magat and HuberViscusi et al. (1991). The effect of a regulation on premature mortality is monetized using the VSL, as discussed in the next subsection.Footnote ²⁰

The direct utility effect of reducing air pollution haze and increasing visibility is another possible benefit of controlling air emissions (arrow 11). The stated preference literature describes the valuation of this difficult-to-quantify benefit, for example Reference Boyle, Paterson and CarsonBoyle et al. (2016).

The right-most branch of the figure shows a way of deriving air pollution-related health benefits that avoids the regular use of CTMs (arrow 12). CTM simulations are time intensive, and technical and/or resource limits constrain the use of CTMs. An alternative is to rely on reduced complexity models that give benefits per unit of emissions reduced. These models use response surfaces, regression analysis, or other methods to link emissions reductions at sources to concentrations at receptors (Reference Gilmore, Heo and MullerGilmore et al., 2019). The changes in concentrations are converted into benefits in the standard way, that is, using exposure-response functions, and valuing the benefits. Relating the benefit estimates to the emissions reductions gives a benefit per unit multiplier. Once these multipliers are compiled, they can be directly applied to source-specific emissions reductions to give receptor-area benefits.

Figure 1 indicates that AEC regulations generate a multiplicity of benefits. An additional channel for co-benefits is related to the technical fact that pollution control technologies can reduce more than one pollutant. For example, baghouse filters entrain pollutants based on size, regardless of type or chemical composition. Regulations targeting one kind of pollutant, for example, mercury emissions, can also reduce others, such as PM2.5. Different pollutants present different health risks, and to the extent practical, the benefits of all emissions reductions should be monetized (Reference Driscoll, Buonocore and LevyDriscoll et al., 2015; Reference Markandya, Sampedro and SmithMarkandya et al., 2018). However, itemizing co-benefits can lead to double counting if nonexclusive valuation methods independently generate the benefit estimates. Thus, care needs to be taken in the aggregation of co-benefits.

Quantifying and measuring this diverse array of benefits poses methodology and data challenges. Given space constraints, we focus on just two topics in the remainder of this section that have generated significant attention in the literature: (1) the value of reducing the risks of illness and premature death, and (2) the value of reducing greenhouse gas emissions.

4.3 Economic Value of Reducing Risk of Illness and Premature Death

We begin with the harder case of lifesaving benefit, and then consider the seemingly more tractable case where the frequency (or severity) of illness is reduced, without any impact on longevity.

Beginning with the insights of Reference Schelling and ChaseSchelling (1968) and Reference Jones-LeeJones-Lee (1976), the approach has been to determine the willingness to pay (WTP) of the affected public for the risk reductions, and those WTP estimates become the primary economic measure of the health benefit for use in BCA. The risk evaluation does not resemble WTP for saving an identified life, such as a trapped coal miner or a patient in urgent need of kidney dialysis. Instead, analysts need estimates of the WTP to protect anonymous citizens that will experience statistical reductions in their probabilities of death and serious illness when exposures to pollution decline.

It turns out that people are willing to pay a substantial amount for such a seemingly small reduction in annual mortality risk, for example, to reduce the average risk of PM-related death from 8 chances in 100,000 to 7 chances in 100,000 per year, or an incremental reduction of 1 chance in 100,000 per year. In the United States, those values have played an important role in making an economic case for environmental regulation (Reference GrahamGraham, 2008).

Two approaches have been used to measure the public demand for risk reduction in monetary units. First, economists measure the revealed preferences of workers and consumers when they confront mortality risks in daily decisions. This literature, which has exploded worldwide in the last forty years, finds that avoidance of a 10(−5) annual mortality risk is valued, on average, at $100 per year in the United States and by smaller amounts in countries with lower average incomes (Reference Hammitt and RobinsonHammitt and Robinson, 2011; Reference ViscusiViscusi, 2018) Second, economists measure the stated preferences of respondents who are asked questions that entail making tradeoffs between money and mortality risks. The results of the stated preference studies also suggest significant WTP for mortality risk reductions but the magnitudes of the estimates may be significantly smaller than the revealed preference estimates (Reference Navrud and LindhjemNavrud and Lindhjem, 2011).

The VSL is a summary measure used to describe estimates of the economic value of lifesaving. Suppose a population of 100,000 individuals is each exposed to a 10(−5) mortality risk. Risk analysts say that 1 “statistical life” is at stake. A VSL value of $10 million implies that the 100,000 individuals, together, are willing to pay $10 million to avert the statistical death, or an average of $100 per individual. In the United States, federal regulatory agencies are now using VSL estimates around $10 million while some European countries tend to use smaller VSL values from $1 to $5 million. The smaller VSL values outside the United States reflect at least two factors: smaller per capita incomes and greater reliance on the stated preference method of valuation.

The current estimate of VSL used in the United States is based on the mean of a fitted distribution of twenty-six estimates of VSL from studies conducted between 1974 and 1991. Five of the studies are stated preference studies and twenty-one are hedonic wage studies. The EPA is in the process of determining whether a revision of this estimate is needed. There is continuing research on the value of the VSL, updating past studies or applying new methods (see Reference Robinson and HammittRobinson and Hammitt, 2016).

One question is whether VSL should be a constant value or vary according to the disease context and the characteristics of the population. Research suggests that the value is heterogeneous (Reference Greenberg, Greenstone, Ryan and YankovichGreenberg et al., 2021; Reference ViscusiViscusi, 2010). The economics literature supplies evidence that VSL varies by cause of death (e.g., higher for cancer than traumatic injury), by age of the affected population (e.g., high for children, highest in the middle of the lifespan, and declining slowly after the age of sixty-five), by income and wealth, and by other factors. However, regulatory agencies tend to use uniform VSL values, perhaps fearing that valuing some subgroups less than others raises sensitive ethical and political issues. In an environmental context where pollution reduction may extend life only slightly toward the end of the lifespan, some analysts argue for a different metric: the value of statistical life year (VSLY). The VSLY may still be quite large, say $300,000 per year in the United States, since elderly citizens may value their remaining years of life highly and may have accumulated substantial assets to support their WTP for additional longevity.

One of the reasons that VSL is not a constant is that some ways of dying are associated with a longer period of morbidity, and more pain and suffering, than others. The revealed preference literature suggests that this period of suffering prior to death is not unimportant but the majority of the VSL value is attributable to the fact that life ends prematurely (Reference Gentry and ViscusiGentry and Viscusi, 2016).

Some environmental regulations reduce the risk of morbidity (e.g., bouts of bronchitis from inhalation of air pollutants) without changing mortality probabilities. Different values of statistical injury (VSI) are available for illnesses of different durations and severity (Reference CameronCameron, 2014). For a nonfatal illness that is associated with an extended hospital stay and some posthospital recovery time, the average VSI may be as high as $300,000 to $500,000 per case, or about 3 percent to 5 percent of the average VSL value (Reference Gentry and ViscusiGentry and Viscusi, 2016).

Many of the benefits of reducing nonfatal illnesses are experienced by the people affected by the illness. They will presumably account for those private impacts in either their stated or revealed preference measures. There are also external costs of nonfatal illnesses that can be substantial. They include foregone production that disrupts employers and payments for health care resources that are made by the government or by private insurers with imperfect risk pools. Some of these external impacts also occur with premature death and ideally should be included in a full accounting of health benefits (UK, 2016).

4.4 Pricing Greenhouse Gas Emissions

Pricing greenhouse gas emissions is commonplace in OECD countries. There are two basic methods: the SCC and the TCP. The SCC is the discounted present value of the stream of damage resulting from this period’s emission of one metric ton of carbon (or CO₂). The SCC is used for regulatory BCA in the United States, Canada, and Germany (US GAO, 2020). In contrast, the TCP is the cost of reducing the marginal ton of carbon along a trajectory to a carbon emissions target, or global temperature limitation. The TCP is used in France and, since 2009, in the United Kingdom (US GAO, 2020). The TCP approach is consistent with implementation of a least-cost strategy for achieving the targets specified in the Paris Climate Accord.

5.1 Introduction

The social discount rate (SDR) is the threshold rate of return required to justify public projects or policies. It is used to generate weights (“discount factors”) that decline over time.Footnote ²⁷ Applying these weights converts future benefits and cost streams into their present value equivalents. Present value equivalents are commensurable with initial period costs, enabling a temporally consistent benefit-cost comparison.

Two SDR concepts are discussed in the discounting literature, the social opportunity cost of capital (SOC) and the consumption discount rate (CDR). The SOC represents the private rate of return foregone when a policy displaces investment. The CDR is the consumer’s marginal willingness to trade current for future consumption; it measures intertemporal opportunity costs when a policy displaces consumption. The CDR can be derived from market data or using the Ramsey discounting formula.

As is commonly pointed out, different discounting approaches converge to the same rate in an economy with perfect capital markets. In actual markets, risk premia, taxes on capital returns, and taxes on the interest earnings of savers create wedges between pre- and after-tax returns. Credit rationing and risks also differentiate credit market segments. These market characteristics give rise to different discount rates, requiring a choice about the appropriate discount rate for public policy.

Discounting methodology must also address uncertainty of various kinds. These include nondiversifiable systematic risks affecting regulatory performance, such as positive correlations between benefits and economic growth, as well as uncertainties about economic growth processes, discount rates, and rare catastrophic events (Reference GollierGollier, 2013; Reference Groom, Drupp, Freeman and NesjeGroom et al., 2022). Unless otherwise noted, this section abstracts from systematic risks to consider other types of uncertainty.

Uncertainties are particularly pertinent for climate regulations. Capital markets provide diminishing information as time horizons increase,Footnote ²⁸ long-range forecasts have limited precision, and intergenerational tradeoffs raise philosophical questions. Moreover, exponential discounting greatly reduces the present values of longer-term returns. For example, the DICE model estimates that the 2025 value for the SCC drops from $140/ton to $22.6/ton as the discount rate rises from 2.5 percent to 5 percent (Reference NordhausNordhaus, 2017). These issues make the choice of intergenerational discount rates especially challenging.

We now turn to the main topics of this section.Footnote ²⁹ The next two subsections describe the market-based and Ramsey rule derivations of the CDR, before turning to the SOC discounting method. The next subsection addresses the discounting implications of the incidence of regulations on consumer credit markets. The final subsection describes the appropriate discount rate when intragenerational efficiency and intergenerational equity are separated as policy objectives.

5.2 The Market-Based CDR

Savers who invest in low-risk government securities receive a relatively stable after-tax rate of return. This return compensates the investor for the value of current consumption foregone and is commonly used as the CDR.Footnote ³⁰ Capital taxes and taxes on interest paid to savers vary throughout the OECD, giving different after-tax returns (hence, different CDRs) on low-risk government securities.

The size and liquidity of bond markets provide the empirical benefit of reliable data (CEA, 2017). However, a limitation of using bond markets as the basis for the CDR is that a large fraction of the public does not hold government debt.Footnote ³¹

Rates of return on government securities have a term structure. Returns on a ten-year treasury bond are the basis for the 3 percent CDR recommended in OMB Circular A-4 in the United States (OMB, 2003). Real bond yields in OECD countries have declined in the last two decades. In the United States, the real yield on ten-year treasury notes is beneath 2 percent (CEA, 2017). The decline in real bond rates has led to recommendations to lower the CDR for US regulatory evaluation (Reference Carleton and GreenstoneCarleton and Greenstone, 2021).

The CDR is recommended in the guidance documents of the European Union and a number of European countries (see Reference FlorioFlorio, 2014; Reference Groom, Drupp, Freeman and NesjeGroom et al., 2022). These CDRs are calibrated from rates earned on government securities but also using the Ramsey discounting rule. Consumption discount rates range between 1 percent (Germany) and 4 percent (Ireland). Like the United States, Denmark uses both the CDR and the SOC as the basis for regulatory evaluation.

CDRs in the United Kingdom and European countries frequently decline as step functions over time (Danish Finance Ministry, 2019; Reference Groom, Drupp, Freeman and NesjeGroom et al., 2022). Circular A-4 in the United States allows for a lower, constant discount rate for regulations having intergenerational effects. Intergenerational welfare is a common motivation for low or declining discount rates. However, “certainty-equivalent” discount rates decline over time when the discount rate itself is uncertain. This occurs when discount rates are constant over time but uncertain, or when the discount rate is not fixed over time, but the stochastic process perpetuates disturbances (Reference Newell and PizerNewell and Pizer, 2003).

The certainty-equivalent rate can be computed as an average over a time horizon or as an instantaneous rate at a given time.Footnote ³² To illustrate, assume there is an equal probability of a 3 percent or 7 percent discount rate in all time periods. In this case, the average and instantaneous certainty-equivalent rates at fifty years are 4.13 percent and 3.48 percent, respectively. At 100 years, the corresponding rates are 3.68 percent and 3.07 percent. In the long run, the certainty-equivalent rate declines to the lower bound of the distribution. The convexity of the discounting factor, and a property described by Jensen’s inequality, account for this result (Reference Cropper, Freeman, Groom and PizerCropper et al., 2014).

5.3 The Ramsey Discounting Approach

The Ramsey discount rate is used in BCA in the United Kingdom and a number of European countries (Reference Groom, Drupp, Freeman and NesjeGroom et al., 2022). It is also used for project evaluation in the EU when financing is sourced from structural adjustment funds (Reference FlorioFlorio, 2014).

The Ramsey rule is based on a utilitarian welfare function. A social planner maximizes intertemporal welfare over an infinite horizon, with utility in each moment t represented as an increasing and strictly concave function of average (per capita) consumption, $U(c_t)$ . The utility function is assumed to be time invariant, and utilities across time are assumed to be additively separable. An isoelastic utility function, $U(c_t)=\frac{c_t{}^{1-\eta }}{1-\eta }$ , is commonly used.Footnote ³³ The term, $\eta =-\frac{c{U}_{cc}}{U_c}>0$ , denotes the absolute value of the elasticity of the marginal utility of consumption, and is constant in the isoelastic specification. The higher the value of $\eta$ , the more concave the utility function. This implies more rapid decline in marginal utility $(c_t{}^{-\eta })$ over time, if consumption is growing, and therefore, an increasing willingness to smooth consumption. It also implies greater aversion to intra-temporal income disparities and, when project outcomes are modeled as uncertain, greater risk aversion.Footnote ³⁴

Given a utility discount rate, $\rho ,$ and growth rate of consumption, $g$ , maximizing intertemporal welfare yields the optimality condition: $r=\rho +\eta g$ . At the margin, this rule equates the rate of return on savings, r, with the rate of return on consumption ( $\rho +\eta g$ ). The expression $\rho +\eta g$ is the standard Ramsey discount rate, providing the basis for calibrating the CDR from the $\rho$ and $\eta g$ components. The term, $\eta g$ , shows the marginal value of consuming now rather than later, given $\eta >0$ and the (default) assumption that $g>0$ .

In the context of intergenerational discounting, the term $\rho +\eta g$ can be interpreted as a normative rule for trading societal welfare over long horizons. In this case, the value for $\rho$ is based on an ethical view about the degree to which future utilities should be discounted – it is commonly set to zero – while the $\eta$ term becomes a distributional weight indicating intertemporal inequality aversion. The $\eta g$ term shows the rate of return required to compensate current consumption sacrifices when future generations are consuming more (assuming $g>0)$ , and the marginal utility of future consumption is declining.

Given its welfare theoretic origins, the Ramsey discount rate is commonly described as “prescriptive” or “normative.” However, the Ramsey rule is interpreted in different ways in the literature. Some scholars suggest that the Ramsey discount rate should be consistent with a riskless, real market rate of interest, for example, Reference WeyantWeyant (2008).Footnote ³⁵ Others argue that the $\rho$ and $\eta$ parameters should be regarded as purely normative and, therefore, that the Ramsey discount rates cannot be expected to correspond to market rates (Reference Sterner and PerssonSterner and Persson, 2008). Still others argue that the parameters of the Ramsey discount rate should be consistent with the public’s preferences. This is the approach taken by the British Government (Reference Groom and MaddisonGroom and Maddison, 2019). Reference Newell, Pizer and PrestNewell et al. (2022) use a hybrid approach recommended in NAS (2017) in which the forecasted values of the CDR are based on Ramsey components that are calibrated to give CDRs consistent with short-term market rates.Footnote ³⁶

The Ramsey rule can be extended in a variety of ways (see Reference GollierGollier, 2013). One is to incorporate uncertainties about expected growth. If the utility function is isoelastic, and logarithmic growth is independent and identically normally distributed, $\ln (\frac{c_t}{c_o})\sim N(\mu ,{\sigma }^2)$ , an extended Ramsey rule can be written as $r=\rho +\eta g\text{'}-\eta (1+\eta ){\sigma }^2$ , with $g\text{'}$ the expected consumption growth rate. The third term introduces “prudence” (denoted by $(1+\eta )$ ) and the volatility of future income, ${\sigma }^2$ , which in combination, lead to precautionary behavior that lowers the discount rate by a constant amount (Reference Gollier and HammittGollier and Hammitt, 2014).

Shocks to economic growth are likely to be positively correlated and, in this case, the persistence of the shocks magnifies future uncertainty, increasing the precautionary saving motive. Using an isoelastic utility function, this causes the discount rate to decline over time. Discount rates will also decline when the growth process is independent and identically distributed, but the parameters of the distribution ( $\mu$ or ${\sigma }^2$ ) are uncertain, or if the $\eta (t)$ parameter declines, rather than remains constant (Reference Gollier and HammittGollier and Hammitt, 2014).Footnote ³⁷

A common modification to the Ramsey rule is to add a small risk premium to the $\rho$ term to account for catastrophic events, such as nuclear wars or natural disasters, that pose an existential risk to humanity. Reference Dietz and SternDietz and Stern (2008) suggest a risk premium of 0.1 percent, representing a 1 in 1,000 annual risk. The British Government adds a 1 percent risk premium to the Ramsey discount rate to cover a combination of catastrophic risk and systematic risk arising from the covariance of project returns and economic growth.

The Ramsey rule has also been extended to account for expected temporal trends in the relative value of environmental services and produced consumption goods. In this context, environmental services and produced consumption goods enter utility, and a modified CDR is derived with the environmental services taken as a parameter in the optimization (Reference Gollier and HammittGollier and Hammitt, 2014). Given standard assumptions about the functional form of utility, the net-return on savings is $r-p$ , where $p$ is the rate at which the value of the environment services changes relative to consumption (Reference Hoel and SternerHoel and Sterner, 2007). This lowers the effective discount rate for $p>0.$

It is commonly assumed that the growth rate of produced consumption goods is higher than environmental services (which may be declining with economic growth). Technical progress extends the supply of produced consumption goods while environmental services are relatively inelastic. It is also assumed that the elasticity of substitution between environmental services and private consumption goods is relatively low, implying that the value of environmental services is increasing over time as they become relatively scarcer.Footnote ³⁸ Making these assumptions, and assuming the relatively high initial discount rate used in the DICE model (5.2 percent), Reference Sterner and PerssonSterner and Persson (2008) show that optimal carbon emissions trajectories are lower beyond about 2080 than those found in the Stern report using a lower initial discount rate (3.4 percent). In other words, relative price changes have a significant impact on the effective discount rate.Footnote ³⁹

An “environmental discount rate” can also be derived with environmental services treated as the choice variable and consumption taken as a parameter (Reference Gollier and HammittGollier and Hammitt, 2014). If there is no substitutability between environmental services and consumption, “dual discount rates” for environmental services and for consumption become necessary (Reference Weikard and ZhuWeikard and Zhu, 2005).

The standard Ramsey discount rate is consistent with the Ramsey/Cass-Koopmans (RCK) economic growth model. The economic growth literature also includes models that treat growth as endogenous, and models that add human capital and environmental resources to production and/or utility functions, among other modifications (e.g., Reference Acemoglu, Aghion, Bursztyn and HemousAcemoglu et al., 2012; Reference Dietz and VenmansDietz and Venmans, 2019; Reference Krutilla and ReuvenyKrutilla and Reuveny 2002, Reference Krutilla and Reuveny2004, Reference Krutilla and Reuveny2006; Reference XepapadeasXepapadeas, 2005). The first order conditions from these models yield differential equations for consumption growth that imply consumption rates of discount. As an example, adding “nature capital” the utility function of a standard RCK model yields an augmented model in which increasing consumption in the present reduces future utility through two channels (Reference Krutilla and ReuvenyKrutilla and Reuveny, 2002). First, as in the standard Ramsey framework, future consumption of the produced consumption good declines with lower investment. Second, environmental amenities decline with less nature capital regenerated, lowering welfare through the direct utility channel. These dual impacts are reflected in a modified Ramsey rule: $r=\rho +\eta g-\frac{V_s}{U_c}.$ The added term, $\frac{V_s}{U_c}$ , is the marginal rate of substitution of the environmental amenity for consumption. Because $\frac{V_s}{U_c}>0,$ the effect of the resource-generated amenity is to lower the Ramsey discount rate; thus, the steady state exhibits a higher level of (nature) capital and a lower level of consumption than in the standard Ramsey model. In contrast to the literature on relative price changes, this result obtains when resources and consumption are perfectly substitutable (the condition assumed in this stylized model), and whatever the change in the growth rates of consumption and natural resources (which are identical in this model along a balanced growth path, giving $\frac{V_s}{U_c}$ as a constant). This is just one example from a large economic growth literature of how the economic context can matter for structuring the Ramsey discount rate.

5.4 The Social Opportunity Cost of Capital

5.4.2 Shadow Price of Capital

The “shadow price of capital” can give the same result as the weighted cost-of-capital method, under some conditions.Footnote ⁴² To illustrate, consider a simple two-period model in which a dollar is invested in an initial period yielding a single consumption benefit, $B,$ in the following period. The social cost of a unit of investment in the initial period can be converted into a consumption numeraire using ${\theta }_o=\alpha \nu +(1-a),$ where ${\theta }_o$ is the shadow price of capital (in present value units of consumption); $\alpha$ and $(1-\alpha )$ are the shares of investment and consumption diverted by a public policy, and $\nu \equiv \frac{1+r}{1+r_c}$ gives the present value, from the consumer perspective, of the rate of return on capital (using r to denote the SOC and $r_c$ to denote the CDR for notational convenience).Footnote ⁴³ Given these definitions, the net present value (NPV) of a dollar of investment using the consumption numeraire is $NPV=-{\theta }_o+\frac{B}{1+r_c}.$ Dividing through by ${\theta }_o$ and simplifying gives the NPV in dollars terms:

$\frac{NPV}{{\theta }_o}=-1+\frac{B}{\alpha (1+r)+(1-\alpha )(1+r_c)}.$

The right-hand side gives the weighted cost-of-capital method. Reference Li and PizerLi and Pizer (2021) show that the consistency of the shadow price of capital method and the weighted cost-of-capital discount rate also holds when future consumption benefits are in the form of an annuity.

In the more general case, the time horizon is variable, future benefits are not in the form of an annuity, and not all benefits are consumed. Reference Li and PizerLi and Pizer (2021) derive a general formula encompassing these conditions:

$SDR=(1+r_c){\left(\frac{{\theta }_o}{{\theta }_t}\right)}^{\frac{1}{t}}-1.$

The new element, ${\theta }_t$ , is the shadow value of benefits in period t given that some fraction of the benefits in each period are reinvested. When ${\theta }_o={\theta }_t$ , the $SDR$ goes to $r_c$ (the $CDR$ ). Reference Li and PizerLi and Pizer (2021) explore a range of other plausible values for $\frac{{\theta }_o}{{\theta }_t}$ . When the time horizon is short, for example, $t<10,$ the ratio $\frac{{\theta }_o}{{\theta }_t}$ dominates the result, but as the time horizon increases, the ratio becomes less significant. For a fifty-year horizon under best estimates for parameter values and a 3 percent CDR, the range of possible SDRs falls to between 2.3 percent and 3.8 percent (Reference Li and PizerLi and Pizer, 2021). As $t\to \infty$ , ${\left(\frac{{\theta }_o}{{\theta }_t}\right)}^{\frac{1}{t}}\to 1,$ and SDR converges to the consumption rate of discount.

Critics argue that the shadow price of capital approach is not practical to implement (Reference HarrisonHarrison, 2010), and that exogenous reinvestment rates are inconsistent with savings behavior (Reference Burgess and ZerbeBurgess and Zerbe, 2011). Reference Li and PizerLi and Pizer (2021) address the later point assuming that the underlying dynamic system is in a steady state, giving constant consumption and savings rates, and derive the discount rate using a Ramsey model.

5.5 On-Budget versus Off-Budget Financing

The positive discounting literature assumes that government budgets finance projects, and the financing is sourced from the capital market.Footnote ⁴⁴ As noted in Section 5.4, the associated financing opportunity costs are additional consumer savings, displaced investments, and borrowing from foreign sources. This paradigm reflects the origins of BCA in infrastructure evaluation. For several reasons, the context for regulatory evaluation is quite different. As noted in Section 3, output adjustments offer one margin for regulatory compliance. For polluters, reducing output provides operational cost savings, rather than additional capital costs. The opportunity costs associated with output reductions – foregone consumption – are entirely borne by consumers. On the inframarginal output range, polluters may bear some of the costs. But except for the case of technology standards, polluters have considerable flexibility for input adjustments that do not require capital. An example is the compliance achieved by US utility plants from fuel-switching to national gas from 2005 to 2015, rather than installing pollution controls. Overall, the capital cost share in total social costs is likely to be significantly lower for regulations than for infrastructure projects.

The SDR using the opportunity cost approach should reflect the impact of government interventions on the marginal trade-off of current for future consumption in the entire economy – an inherently GE concept. In this context, the incidence of regulatory compliance costs is relevant. In GE models using constant returns to scale production technology, all compliance costs are borne by consumers. Specifications which allow upward sloping supply in some sectors allow pass-through rates of less than 100 percent, consistent with the literature on regulatory pass-through. Whatever the degree of pass-through, consumer prices in CGE regulatory cost models increase in response to regulations, affecting real wages and labor supply (see Section 3).

A relevant question is how GE adjustments affect consumer behavior in the credit markets they use. This question has not been explored in the discounting literature. If consumers use of credit markets is impacted by regulation, the opportunity costs are likely to be high. Credit rationing is pervasive in the economy (Reference Stern, Stiglitz and TaylorStern et al., 2022). Several studies have documented high marginal rates of consumer time preference consistent with credit rationing. For example, the choices of US military personnel about compensation packages offered as part of downsizing program revealed implicit nominal discount rates between 10 percent and 54 percent (Reference Warner and PleeterWarner and Pleeter, 2001). A field study in Denmark showed a mean nominal rate of 28 percent (Reference Harrison, Lau and WilliamHarrison et al., 2002). An implicit nominal discount rate of 77 percent and a mean required payback period of 3.7 years were estimated in a revealed preference study of energy investment behavior in Greece (Reference Damigos, Kontogianni, Tourkolias and SkourtosDamigos et al., 2021). Demographic characteristics, such as income and education, affected these estimates, as well as the size of investment costs (Reference Damigos, Kontogianni, Tourkolias and SkourtosDamigos et al., 2021).

RIAs of energy efficiency regulations in the United States have had to confront the disparity between consumer behavior and the economic parameters assumed in the analysis. RIAs commonly show that the market value of energy savings is high enough to pay off private costs, using standard assumptions about SDRs and other economic parameters. Why markets do not independently offer these technologies is a puzzle known as the “energy efficiency gap” or “energy efficiency paradox” (see Reference Gillingham and PalmerGillingham and Palmer (2014) and Reference Helfand and Dorsey-PalmateerHelfand and Dorsey-Palmateer (2015)). Consumer discount rates that are higher than the assumed SDRs is one possible reason for this discrepancy. Bounded rationality and low levels of financial literacy are also possible causes.

The distinctive incidence of regulations, and the possible relevance of GE effects for discount rates in regulatory evaluation, have not been addressed in the discounting literature. Exploring this topic would be a worthy objective for future research.

5.6 Dichotomizing Efficiency and Intergenerational Equity

Using two policy instruments to promote two policy objectives, if the choice of instruments is unconstrained, is a basic optimization principle. Using intergenerational wealth transfers to address intergenerational equity, while undertaking high yield investments to promote economic efficiency, is an application of this principle. Suppose the CDR is 2 percent and the SOC is 8 percent. A $1,000 energy investment yielding a single consumption benefit at 2 percent would give the equivalent of $7,389 in 100 years. However, only $2.48 invested in a sovereign wealth fund yielding 8 percent would give the same result.Footnote ⁴⁵ Or investing $1,000 at 8 percent would give $2,980,942 in 100 years. Pursuing a Pareto improving strategy that provides these gains is the most economically efficient way to address intergenerational concerns (see Reference Burgess and ZerbeBurgess and Zerbe (2011), Reference HarrisonHarrison (2010), Reference LindLind (1995), Reference Weisbach and SunsteinWeisbach and Sunstein (2008), and Reference WeyantWeyant (2008)).Footnote ⁴⁶

An intuitive way to implement this principle is for the current generation to decide upon the magnitude of the intergenerational transfer; for example, the size of the endowment for a sovereign wealth fund. Conditional on this decision, all generations would support investments from the fund that maximize future returns (Reference KaplowKaplow, 2006). Suppose that the fund’s long-run rate of return is expected to be z percent. The fund could be used to finance projects impacting future generations with social returns greater than z percent, or the fund could be allowed to grow at z percent. This context is completely analogous to capital budgeting, with z percent the discount rate for prioritizing projects having long-lasting effects.

This concept seems appealing for the public financing of long-lasting infrastructure, and R&D in basic sciences, early-stage energy development, biotechnology, computer and information systems, and the like. It seems less appealing for addressing environmental issues having uncertain but enduring and irreversible effects – such as the accumulation of atmospheric carbon stocks. In this case, transfers to future generations, either as compensation, or to finance adaptations, might be less than originally assumed to be necessary to meet intergenerational obligations. Moreover, the low substitutability between produced goods and environmental services could render money compensations relatively ineffective.

A better approach is to adopt a budget limitation based on an environmental performance metric, rather than a money metric. Of course, this is exactly the approach of target-consistent carbon pricing. As noted in the previous section, the objective of target-consistent carbon pricing is to direct investments to projects yielding the lowest net costs, consistent with a stipulated emissions trajectory. Although the discounting literature has discussed a SCC-compatible discounting approach (Reference Newell, Pizer and PrestNewell et al., 2022), the discounting implications of the target-consistent carbon pricing method have not been considered, to our knowledge. It is intuitive that the appropriate discount rate is closer to the long-term yield on a sovereign wealth fund than a CDR; the latter will not be stringent enough to sort out more from less efficient projects. Moreover, a lower rate motivated by intergenerational equity concerns is not apt in this context, given that intergenerational concerns are addressed ex ante when the emissions target is established. The discounting implications of the target-consistent carbon pricing deserve additional research, in view of the paucity of academic literature on this subject.

6.1 Introduction

AEC regulations impose uncompensated losses on private actors and offer a mix of market and nonmarket benefits to large populations. The beneficiaries are not necessarily the individuals who bear the costs. This profile makes the distributional effects of regulations both policy relevant and difficult to assess.

Distributional impacts can be categorized along different dimensions, including factor income and employment; position in the market as a producer or consumer; business size, or entry date into the market; jurisdictional level; regional demarcation, or some combination. Benefit-cost analysis guidance documents recognize the importance of distributional effects, and distributional concerns are reflected in policy designs that favor some constituencies (small business, incumbent polluters) over others. However, owing to complexity of the distributional pattern for AEC regulations, distributional assessments in RIAs are often relatively limited. The most common practice is to identify impacts by identifiable stakeholder groups, like small business or labor in regulated industries, when information is available.

AEC regulations are sometimes perceived as having regressive effects on worker incomes and employment. Lower income individuals are disproportionately employed in polluting industries and spend a larger share of income on energy-intensive goods, like electricity and fuel. However, the literature suggests a more nuanced picture. The impacts of regulations on lower income individuals depend on the type of regulation, the structure of the economy, regional environmental effects and demographic patterns, and institutional characteristics, such as the nature of the tax and transfer system. These factors are likely to empirically vary by specific regulatory and economic context.

CGE models can provide income class-disaggregated surplus measures for policy change that embody all welfare channels represented in the model. The income spectrum is typically stratified into quintiles or deciles; however, Reference Rausch, Metcalf and ReillyRauch et al., (2011) develop a CGE model that disaggregates effects on individual households. Welfare effects can also be demarcated by countries or regions within a common economic area, such as the EU (Reference Mayeres and Van RegemorterMayeres and Van Regemorter, 2008)

The distributional literature commonly focuses on one or more aspects of the total welfare effect of a regulation, to better understand the components. This practice drives the structure of this section. Following the expositional format in Reference FullertonFullerton (2011), we start with an assessment of the burden of regulations on producers and production factors. The next topic is the impacts of regulations on consumers. We then consider how the allocation of pollution rents affects distributional impacts. Last, we review the distribution of the health and environmental benefits that AEC regulations provide.

6.2 The Burden of Regulations on the Supply Side

AEC regulations have two sorts of effects on workers, capital owners, and other production factors. First, if the regulation affects relative input prices broadly in the economy, the earnings of factors that remain employed will change. The net-effect depends on the share of income received from different sources and the impact of the regulation on their returns. Second, the incentive effect of AEC regulations will alter the allocation of inputs in the economy, giving rise to both efficiency costs and distributional effects. As discussed in Section 3, economic structure and regulatory design influence the burden of the regulation on the supply and demand sides, driving the extent of these changes.

The relative burden of government interventions on labor and capital is widely studied. The classic research on corporate tax incidence by Reference HarbergerHarberger (1962) shows that results are extremely sensitive to the parameterizations of utility and production functions. These results carry over in extensions that add emissions into the production function and analyze the burdens of an emissions tax (Reference Fullerton and HeutelFullerton and Heutel, 2007) and a standard limiting emissions per unit of output (Reference Fullerton and HeutelFullerton and Heutel, 2010). Whether emissions restrictions relatively burden capital or labor more heavily is quite ambiguous in these models.

Harberger-type models represent the boundary case where labor and capital are in fixed supply. Another limiting case is when supply of capital is freely mobile at a fixed world rate of return. In this situation, the burden of an environmental tax falls exclusively on labor and consumers (Reference Fullerton and MuehleggerFullerton and Muehlegger, 2019).

Production factors in these models are fully employed in equilibrium, and adjustment costs are not fully represented. Unemployment and adjustment costs are often politically salient. Concerns about asset-specific capital losses that arose in the United States in the era of utility restructuring have again resurfaced in the context of climate regulations, which are likely to accelerate the retirements of fossil-fuel plants (Reference CaldecottCaldecott, 2018). Workers whose skills are closely tied to their current use or location have limited adjustment flexibility and may become involuntarily unemployed. To the extent that sector-specific labor unemployment is long-lasting, the personal and societal costs are well established (Reference BartikBartik, 2015; Reference Haveman and WeimerHaveman and Weimer, 2015). If such employment effects can be accurately measured and attributed to AEC regulation, the social costs are relevant for both the distributional and efficiency analysis.

Labor that is not occupation or sector-specific can relocate at a relatively low transition cost without a significant change in earnings. Even in this case, however, welfare losses are likely. An equilibrium sorting model by Reference Kuminoff, Schoellman and TimminsKuminoff et al. (2015) demonstrates why. In this model, individuals make simultaneous choices about employment and living locations. Changing locations affects employment possibilities and income, but also other attributes, such as commuting distances, housing choices, and community characteristics. Given that these other attributes affect welfare, economic estimates of involuntary relocations based on earnings differences alone can understate welfare losses.

Partial equilibrium models are sometimes used to assess the net employment impact of environmental regulation on regulated sectors. The effects on sectoral output prices are incorporated into these models, but factor prices and other commodity prices are typically held constant. Using this approach, Reference Belova, Gray, Linn, Morgenstern and PizerBelova et al. (2015) show that three factors influence the employment impact of a regulation. First, holding labor per unit of output constant, labor usage will decline from the output substitution effect discussed in Section 3. Secondly, and again holding labor intensity constant, labor usage will increase as a function of the cost increase the regulation brings about. Third, the regulation can change labor intensity, increasing or decreasing the share of labor in total cost. With the first two factors having opposite impacts on employment, and the third directionally ambiguous – depending on whether labor is a substitute or complement with the regulated input – a regulation’s overall effect on sectoral employment is ambiguous.

The employment effect of regulations economy-wide has been studied in CGE models that incorporate labor market processes with equilibrium unemployment. Union bargaining is one explanation for the persistence of involuntary unemployment. Union bargaining models have been used to explain labor market behavior in Europe, where unions are relatively influential. A “right-to-manage” model is an example. In this framework, trade unions and employer representatives negotiate over a wage rate. After the wage is determined, firms choose the employment level that maximize profits (Reference KoskelaKoskela, 2001).

The employment effects of “green tax reforms” have been studied in analytical GE models that incorporate a union bargaining process. The tax reform consists of shifting taxes from labor to dirty polluting goods or energy inputs while maintaining budget balance. The employment effect of tax shifting among consumption goods is found to be ambiguous (Reference Koskela and SchöbKoskela and Schob, 1999). Institutional characteristics, such as whether unemployment benefits are indexed to the price level, affects the results. Shifting taxes from labor to energy can increase the equilibrium employment level if labor is substitutable enough with energy in production (Reference Koskela, Schöb and SinnKoskela et al., 1998).

Search-friction models are another way to model equilibrium unemployment (Reference Hafstead and Williams IIIHafstead and Williams (2020)). In this setup, firms determine the supply of vacancies and incur a cost to post vacancy announcements. Job searching imposes transactions costs on prospective workers. The employment success rate is a function of the number of job searches and vacancies, with the probability of securing employment less than one. Labor market frictions create a gap between the marginal product of labor and a jobseekers’ time opportunity costs. Once a job is located, bargaining over the wage share of the surplus – often modeled using a Nash assumption– is necessary before the employment process is finalized.Footnote ⁴⁷

An analytical model by Reference Bovenberg and Van Der PloegBovenberg and van der Ploeg (1998) incorporate a search-friction labor market to study the effect of a revenue-neutral shift from labor to energy taxes. This model represents both informal and formal sectors. Tax shifting has ambiguous effects on the equilibrium employment level, depending on economic factors (whether income in the informal sector is linked to productivity in the formal sector) and institutional factors (how unemployment benefits are indexed).

Reference Hafstead and Williams IIIHafstead and Williams (2020) study the sectoral and economy-wide employment impacts of environmental policies in a CGE model with a search-friction type labor market. As is common in CGE models, this model captures the employment effects of upstream and downstream linkages, and the effects of consumer demand shifts from changes in product prices. The search-friction labor market adds an additional adjustment channel, involving the matching process between job seekers and prospective employers and the bargaining process over wages. The study considers three types of environmental policies. The first is a sector-specific regulation, modeled as an input cost shock, that targets durable manufacturing. This is taken to represent a technology-based standard. The second is a tradable performance standard implemented in the utility sector that limits CO₂ emissions per megawatt hour generated over the entire industry. The final policy is an economy-wide carbon tax. In all cases, labor taxes are adjusted to maintain budget balance. The policies differ in magnitude (which is not the principal focus of the study). The results are presented for the employment effect in the regulated sectors, the sectors that contract from the regulation (“negative spillover sectors”), those that expand (“positive spillover sectors”), and the sum for the total economy (see Table 1). In all cases, the employment effect in the regulated sector provides a biased picture of the economy-wide effect, and labor reallocations among sectors are significant. A durable manufacturing regulation decreases net employment in the regulated sector, owing to an output substitution effect that reduces labor demand. The net of spillovers in other sectors is to increase economy-wide losses further (see Table 1). In contrast, the power-sector performance standard increases net employment in the regulated sector. Labor is substituted for energy and the rate-based structure of the standard affords an implicit output subsidy that helps to maintain employment. However, the net of spillover effects in other sectors leads to a negative economy-wide employment impact. The carbon tax has a negative employment impact on energy industries, but the net of spillover effects is to increase employment economy-wide, with the revenue recycling mechanism playing a role in this outcome (Reference Hafstead and Williams IIIHafstead and Williams, 2020). The nuances of this distributional picture suggest the value of disaggregating the incidence of employment impacts beyond the regulated market.

Table 1Employment effects of environmental regulation (percent of total employment)

Source: Adapted from Reference Hafstead and Williams IIIHafstead and Williams (2020).

6.3 The Burden of Regulations on Consumers

Some of the cost of regulations is likely to be passed on to consumers. The burden is purely distributional for inframarginal units of consumption, while conventionally efficiency and adjustment costs arise on the consumption foregone. In either case, the regulation most significantly burdens consumers having preregulation expenditures weighted relatively heavily in goods that increase in price, such as electricity or fuel. A common question is whether this burden falls more heavily on low-income consumers than those of other income classes.

Reference Flues and ThomasFlues and Thomas (2015) review consumer expenditure data in OECD countries to assess the burden of energy taxes on different income classes and types of households. As is typical of expenditure-based simulation studies, the study considers the purely distributional effect on inframarginal units of consumption as the price of energy rises. The study finds that the burden of energy taxes varies by the type of the fuel. Comparing these burdens against total consumer expenditures, electricity taxes among OECD countries are generally regressive. Expenditure share on electricity declines as total consumption expenditures increase. Taxes on heating fuels are shown to be mildly regressive, while taxes on transportation fuels are progressive on a 21-country average but vary significantly by country. For example, transportation fuel taxes are progressive in countries with lower per capita GDPs but are proportional or mildly regressive in countries with higher per capita GDPs (Reference Flues and ThomasFlues and Thomas, 2015). The result is attributable to lower automobile usage among lower income households in poorer countries. These results are generally consistent with a review by Reference Pizer and SextonPizer and Sexton (2019), based on similar methodology.

In another study, Reference Cronin, Fullerton and SextonCronin et al. (2019) use expenditure data to show that a carbon tax imposed in the United States is somewhat progressive. While the share of carbon emissions as a fraction of consumer expenditure is relatively constant across income classes, lower income consumers receive a larger fraction of their income as transfer payments. These payments are indexed to the price level, so transfer payments increase when a carbon tax is imposed and prices increase.

Reference Davis and KnittelDavis and Knittle (2019) study the burden of Corporate Average Fuel Economy standards in the United States. These standards provide an implicit subsidy to smaller fuel-efficient cars and an implicit tax on larger cars. Because higher income families buy a relatively larger share of new cars, the standards are mildly progressive when only new vehicle sales are considered. However, the standard also has the effect of shifting some demand to the used car market, thereby raising used car prices. Including the used car market makes the standards mildly regressive, given that low-income consumers disproportionately buy used cars. In either case, the effects are not large compared to average annual household income.

Reference LevinsonLevinson (2019) compares the distributional properties of energy efficiency regulations versus emissions taxes. Consumers pay for efficiency in the purchase price of appliances with differing energy efficiency ratings, and for the usage level through expenditures on energy to operate the appliances. Lower income consumers ordinarily buy older less efficient appliances and run them less than higher income consumers. This behavior translates into a lower level of energy consumption in lower income households. An efficiency regulation forces consumers to buy efficiency. This is relatively less burdensome for richer households who are already purchasing energy efficient appliances. An energy tax burdens both lower and higher income households. But rich households pay more owing to their higher energy consumption levels. Given these differing effects, it turns out that both appliance efficiency standards and energy taxes are regressive, but energy taxes are less regressive than efficiency standards.

Horizontal distributional effects are also increasingly studied in the literature. Heating and cooling loads, modal transportation choices, commuting patterns and travel distances, and the composition and price of energy sources differ within and across regions. These heterogeneities introduce differences in end-use energy loads, holding incomes constant. Household characteristics, including family size and age, also affect energy consumption. In the United States and the United Kingdom, these factors cause variation in energy consumption within income classes that can equal or exceed variation across them, depending on the energy source (Reference Pizer and SextonPizer and Sexton, 2019). For the bottom 40 percent of the consumption distribution, Reference Cronin, Fullerton and SextonCronin et al. (2019) find that the burden of a carbon tax varies more widely among consumers within income classes than between the averages for the lowest and highest deciles. In a static computable GE model that represents all channels affecting distributional burdens, the burden of a $20 carbon tax also varies significantly within income deciles (Reference Rausch, Metcalf and ReillyRauch et al., 2011).

6.4 The Distribution of Pollution Rents

Inframarginal rents on regulated pollution are significantly larger than abatement costs unless the level of regulation is stringent. The magnitude and political salience of these rents, particularly for policies restricting carbon emissions, raise important distributional issues.

One option is to distribute some rents to polluters to compensate (or partially compensate) for abatement costs. When abatement costs can be significantly passed onto consumers, a relatively small share of pollution rents is needed for compensation. A CGE study by Reference Bovenberg and GoulderBovenberg and Goulder (2002) of a $25/ton carbon tax imposed on coal, oil, and natural gas suppliers showed that only 4.3 percent of rents earned from the coal industry and 15 percent from oil and natural gas producers were needed for complete compensation.

Carbon rents can also be used to defray consumer burdens. Rebating revenues lump sum across income can make carbon taxes progress (Reference Rausch, Metcalf and ReillyRauch et al., 2011). However, Reference Cronin, Fullerton and SextonCronin et al. (2019) show that heterogeneities in household characteristics within income classes complicate the compensation of low-income consumers. The institutional structure of the tax and transfer system does not allow the differentiation of transfer schemes to fully reflect household heterogeneities.

A basic constraint is that there are not enough carbon rents to compensate all claimants (Reference Dinan and Holtz-EakinDinan and Holtz-Eakin, 2003). Under these conditions, a large rent-seeking literature shows that the resource costs associated with the rent-seeking activities can attenuate or fully the dissipate rents, for example, Reference Pérez-Castrillo and VerdierPérez-Castrillo and Verdier (1992). An extension to market-based environmental policies demonstrates significant rent-seeking costs (Reference MacKenzie and OhndorfMacKenzie and Ohndorf, 2012). Game-theoretic models of political competition show that the costs associated with policy contestation can exceed abatement costs (Reference Krutilla and AlexeevKrutilla and Alexeev, 2012, Reference Krutilla and Alexeev2014). The key intuition is that conventional efficiency costs are manifest at the margins, while policy-related transaction costs are incentivized over the entire inframarginal and marginal ranges of the regulated activity (Reference Krutilla and KrauseKrutilla and Krause, 2011).

Expanding the menu of policy options could reduce the costs of rent-seeking and political transaction costs. One option is to pair carbon taxes or auctioned tradable permit systems with deregulatory initiatives. Price-based instruments are a substitute for the prescriptive regulations that industries find burdensome. Hence, the option of transitioning from the existing regulatory system to a price-based system would open-up a channel for policy reform. Additionally, the way carbon rents are used should be considered within the context of the full menu of fiscal policy options. It is economically inefficient to treat the allocation of pollution rents as the only policy instrument for achieving distributional objectives, and risks incentivizing high policy transaction costs.

6.5 The Distribution of Benefits

Benefits associated with AEC regulations include energy savings and the value of reducing emissions. The PE distribution of these benefits is considered in this subsection.

Energy savings arise from the use of more efficient consumer appliances or vehicles, or from process adjustments and/or new technology in the commercial or industrial sectors. In the former case, the benefits of energy savings fall on consumers. The benefits of commercial or industrial energy savings are partially borne on the supply side, and partially passed on to consumers as lower prices and/or greater consumption.

Reducing air pollution exposures produces the variety of benefits indicated in Figure 1, Section 4. The benefits of reducing health care costs falls on some combination of health insurers (private and public), health care providers, and consumers, depending on the institutional structure of health insurance systems. The value of improving productivity in renewable resource sectors is distributed among suppliers and consumers of resource-derived products. Increasing labor productivity will be distributed as combination of higher profits for employers and higher wages for workers; consumers experience the benefits of any increases in leisure time. The productivity effects from better conditioned commercial building structures or materials will be shared between the supply and demand sides of the market. The increased value of residential property, and the benefits of reducing haze and improving visibility, are borne by consumers.

Two factors influence the distribution of morbidity and mortality benefits associated with improving air quality. First, regulations reduce exposure risks nonuniformly among regions. Secondly, for a given exposure reduction, marginal valuations vary among individuals (Reference Hsiang, Oliva and WalkerHsiang et al., 2019). One reason marginal valuations are likely to differ across populations is that health or environmental damage associated with air pollution are nonlinear in the level of exposures. If damage functions are convex, marginal damage reductions will decline as air quality improves, translating into declining marginal valuations – all else constant. The standard assumption of diminishing marginal utility independently predicts the same relationship. These factors imply that baseline exposures will influence the marginal value of exposure reductions, and that individuals facing different exposures will be located on different parts of marginal damage functions.

Socio-demographic heterogeneities also give rise to different marginal valuations, holding exposure risk reductions constant (Reference Hsiang, Oliva and WalkerHsiang et al., 2019). Population characteristics, such as age, health status, and ethnicity, are likely to influence the health effects of air pollution, for example, Reference Chay and GreenstoneChay and Greenstone (2003). Preference parameters vary among individuals, for example, parents may have relatively strong preferences for better air quality out of concern for their children (Reference GraingerGrainger, 2012). Relatively health-conscious consumers are more likely to take averting actions, like wearing face masks, staying indoors on a polluted day, or purchasing air conditions or filtration equipment – all else constant. These actions affect valuations and should be incorporated into WTP estimates (Reference Deschenes, Greenstone and ShapiroDeschenes et al., 2017). The effect of income is another source of variability. Higher income individuals should have relatively high marginal valuations – again holding all else constant – if air quality is a normal good.

Generalizing about the distributional impacts of these factors is challenging. For example, the evidence suggests that lower income populations and ethnic minorities experience higher exposure risks than other demographics (Reference Jbaily, Zhou and LiuJbaily et al., 2022; Reference Tessum, Paolella and ChamblissTessum et al., 2021; Reference Thind, Tessum, Azevedo and MarshallThind et al., 2019). All else constant, this would position lower income individuals on higher points on the marginal damage functions. But lower income individuals also are likely to have lower WTPs, making the combined effect ambiguous.Footnote ⁴⁸

It is hard to empirically sort out these effects, and differences in marginal valuations are ignored in RIA.Footnote ⁴⁹ One concentration-response (C-R) coefficient is used to represent a specific health effect of a particular pollutant – for example, the effect of PM2.5 on all-cause mortality – without differentiation for population characteristics (Reference LevyLevy, 2021). Population-averaged unit values, such as the VSL, are taken to represent marginal values. The consequence is that reducing exposure risks nonuniformly is the only channel through which distributional effects on the benefit side can be differentiated in current RIA practice.Footnote ⁵⁰ In fact, some evidence suggests that regulations do reduce exposure risks nonuniformly. Reference Bento, Freedman and LangBento et al. (2015) found that enforcement of the 1990 Clean Air Act amendments targeted polluters in areas closest to pollution monitors in nonattainment regions. These areas have relatively large populations of lower income homeowners.

The incidence of benefits can be realized as changes in property valuations – the basis for the large benefit valuation literature using hedonic models. Measuring changes in property values, Reference Bento, Freedman and LangBento et al. (2015) found that the annual benefits of PM10 reductions from the 1990 Clean Air Act amendments are about twice as high as a fraction of annual income for lowest quintile of homeowners than for the highest, owing to the location of low-income homeowners in polluted areas, and the larger impact of regulation in these areas. For rental properties, the incidence of capitalized values is distributed between renters and landlords. A study by Reference GraingerGrainger (2012) found that about half of the capitalized value of local air quality improvements was captured in higher rental rates. Reference Bento, Freedman and LangBento et al. (2015) found that rental rates were not significantly affected by changes in air pollution except for units located relatively close to air pollution monitors.

The distribution of economic damage from global climate damage is a function of the magnitude of physical climate effects and their marginal valuation (Reference Hsiang, Oliva and WalkerHsiang et al., 2019). Local and regional climate effects, such as precipitation patterns, hydrological conditions, storm frequency and intensity, are heterogenous across regions. Damage functions are likely to be nonlinear and socioeconomic conditions vary widely for different populations. In a review of the literature, Reference Hsiang, Oliva and WalkerHsiang et al. (2019) conclude that physical climate effects are not necessarily greater for lower income populations, but that marginal damage are likely to be higher owing to demographic heterogeneities and nonlinearities in damage functions (Reference Hsiang, Oliva and WalkerHsiang et al., 2019). In the literature on natural disasters, low-income populations are found to be particularly vulnerable to climate change, for example, Reference Hallegatte, Vogt-Schilb, Rozenberg, Bangalore and BeaudetHallegatte et al. (2020).

7.1 Introduction

The economic evaluation of AEC regulations confronts numerous uncertainties that arise in the following areas: (1) the forecast of economic conditions, energy usage, and emissions trajectories in the counterfactual state of the world without the regulation; (2) the effect of the regulatory intervention on energy usage and emissions trajectories; (3) the effect of changes in emissions trajectories on ambient concentrations; (4) the effects of ambient concentrations on exposures, health, and environmental damages; and (5) the valuation of damages. The costs associated the regulatory intervention are also uncertain, as noted in Section 3. Credibly addressing these uncertainties poses an analytical challenge.

In an ideal world, uncertainties would be identified and addressed at every stage of regulatory development and assessment. At the earliest stages of regulatory design, an evaluation would be made of modifications of the structure of the regulation or additional information that reduces uncertainties. At the analysis stage, regulatory alternatives would be systematically compared to assess how uncertainties affect the NPV comparison. At the reporting stage, the regulatory BCA would indicate the level of confidence about the results of the economic analysis, with transparency about the mechanisms that cause the uncertainties. In practice, the uncertainty analysis in RIAs can significantly depart from this ideal.

This section starts with a taxonomy of the levels of uncertainty that can arise in the BCA of AEC regulations. Uncertainty evaluation methods are then considered.

7.2 Uncertainty Levels

How to taxonomize uncertainty is the subject of a large decision-science literature. The classical work by Reference KnightKnight (1921) defines “risk” as the case that parameter values have known outcomes with known probabilities; “uncertainty” as the case that outcomes are known but their probabilities are unknown, and “ignorance” as the situation when even possible outcomes are unknown. Not surprisingly, researchers have proposed numerous additional taxonomies in the more than 100 years following Knight’s seminal work.Footnote ⁵¹ Recent literature parses the distinctions along the uncertainty-to-ignorance part of Knight’s spectrum, using such terminology as “deep” or “fundamental” uncertainty. Reference Walker, Lempert, Kwakkel, Gass and FuWalker et al. (2013) define “Level 4” uncertainty as the context where the probability of outcomes cannot be quantitatively specified but can still be ranked according to their likelihood, while “Level 5” uncertainty denotes more fundamental uncertainty approximating “ignorance” in Knight’s definition. Reference Lempert, Popper and GrovesLempert et al. (2013) define fundamental uncertainty as a situation where “the parties to a decision cannot agree upon (1) the appropriate models to describe interactions among system variables, (2) the probability distributions to represent uncertainty about key parameters in the models, and/or (3) how to value the desirability of alternative outcomes.” For our purposes, “less fundamental uncertainty” is defined as the situation that models are reasonably well-accepted, and parameter values are locally uncertain within “reasonably small ranges,” while “more fundamental uncertainty” is the case that researchers have alternative views and significant disagreements about the structure of models and the value of parameters. We now turn to methods that can be used to address “less fundamental uncertainty,” before considering decision-science approaches that are relevant for incorporating “more fundamental uncertainty” into the evaluation.

7.3 Methods for Addressing Parameter Uncertainty in Reasonably Well-Accepted Models

When uncertainty is less fundamental, deterministic sensitivity analyses that vary parameters locally within a reasonably small range can be used to represent possible outcomes. If probability distributions over uncertainty parameters are known, uncertainty propagation methods such as Monte Carlo simulation become relevant (Reference Morgan, Henrion and SmallMorgan et al., 1990). In this context, value of information approaches such as “real options” can also be used by regulators to assess how changing the timing or scale of the regulation will improve expected economic returns (Reference FarrowFarrow, 2004).

7.4 Methods When Uncertainty Is More Fundamental

When uncertainty is more fundamental, two general methods can be used to support decision-making. Research synthesis summarizes knowledge from evolving research or expert judgments about uncertainties such as the health effects of PM2.5 or the risks of catastrophic climate change. A variety of precautionary decision criteria and/or modeling methods can also be used to inform decision-making. These approaches are discussed in the next section.

8.1 RIA Evolution

The integration of BCA into regulatory evaluation has varied in different parts of the OECD, reflecting different proportionality requirements for the use of RIA, and different traditions for using environmental BCA to evaluate regulations and policies. The use of regulatory BCA seems likely to increase with additional policymaking in the air, energy, and climate policy areas, given the significance of the resource tradeoffs. Methods will continue to evolve as the demand for regulatory analysis continues to grow.

8.2 Cost Analysis

Methods for regulatory cost estimation differ in the degree to which regulatory characteristics, market behavior, and market distortions are represented. Engineering cost data are used to specify the abatement costs of differentiated regulations having a relatively limited impact on the overall economy, and engineering cost estimation remains the modal method in RIA. Regulations that have a significant sectoral or economy-wide impact can justify the resources for constructing partial or GE models that explicitly represent the effects of market adjustments and economic distortions.

The welfare costs of GE feedbacks with market distortions are well documented in the literature. The welfare costs of tax interactions in perfectly competitive models dominate the research; the effects of imperfectly competitive market structures are also studied. Tax interactions and imperfect competition do not necessarily affect regulatory cost estimates in the same direction. In general, the selectivity of model specifications and the focus on particular distortions gives an incomplete picture of the direction and magnitude of the totality of GE effects that could plausibly arise from more general specifications in economies with multiple distortions. Existing research suggests that adding environmental quality nonseparably in utility, adding environmental quality to production, and modeling imperfect labor markets, terms-of-trade effects, and externalities in nonregulated sectors, among other distortions, would help fill in the GE picture. Given the challenges of modeling these factors simultaneously, research on the most high-valued modeling improvements would likely to be useful.

For the most part, the literature on regulatory cost estimation does not address implementation frictions, transactions costs, or the costs associated with rent-seeking and/or policy contestation. Inframarginal rents on regulated pollution are significantly larger than abatement costs unless the level of regulation is stringent; thus, public choice models often show that rent-seeking and political competition impose larger costs than conventionally-measured efficiency costs that are manifest at the margins. Rent-seeking and contestation costs are particularly relevant for carbon restrictions, which have visible impacts on energy prices, and generate sizable rents with observable allocation tradeoffs. The implications of transaction costs and policy contestation on regulatory cost estimates would be useful subjects for future research.

8.3 Benefits

In RIA practice, the benefit and costs of AEC regulations are separately estimated, and market prices, market price-derived shadow prices, or benefit transfers like the VSL are used to value outcomes. Representing preferences for health or environmental quality nonseparably in utility functions in a CGE model, and adding environmental inputs into production functions, would allow the integration of BCA within a conceptually consistent GE framework. Policy-relevant spatial heterogeneities, such as local exposure risks and demographic characteristics, would also need to be represented in this model. Making these adjustments comprehensively enough to be relevant for the evaluation of many kinds of regulations, however, is likely to be difficult for routine practice for the foreseeable future.

The value of reducing mortality risks (VSL) and the price of carbon are two major subjects of debate in the benefit valuation literature. The efficiency and equity implications of disaggregating VSL estimates by income, age, and other attributes are important, but commonly elided in RIA practice due to philosophical and political challenges. Regarding carbon pricing, two valuation methods represent different views on the appropriate role of economic evaluation for supporting climate policy. Using the SCC implies confidence in the capacity of IAMs to forecast the value of long-run climate damages, the belief that discounting is appropriate for representing intergenerational tradeoffs, and the view that BCA should be used to determine the level of greenhouse gas emissions. Proponents of the target-consistent approach are more likely to doubt the reliability of long-run damage valuations, believe that safe margins, in the form of precautionary emissions targets, are normatively justified, and support a deliberative process for establishing carbon emissions targets to address intergenerational equity concerns. In this framework, economic evaluation plays the role of cost-effectiveness analysis that guides the lowest-cost emissions trajectory consistent with the policy goal.

8.4 Discounting

The SOC and the CDR are commonly recommended. The CDR is commonly specified using a riskless rate from the bond market or derived from a social welfare function consistent with a Ramsey economic growth model. The SOC is taken as the real before-tax return on capital and can be derived from national income accounts, giving a rate that embodies an economy-wide average risk premium.

Discounting perspectives in the academic literature are reflected in the guidance documents of different countries. European governments tend to use CDRs with riskless rates in the 1 percent to 3 percent range. Risk premia are often added to these rates. Canada and Latin American Countries usually use SOC-based discount rates, varying from 6 percent to 10 percent. In the past, the United States has used a 3 percent CDR and a 7 percent SOC, but discount rate guidance is under review at the time of this writing.

The intergenerational discounting literature shows that uncertainties about the discount rate or the rate of economic growth can cause declining discount rates over time. The discount rates in European guidance documents decline as step functions over successive time intervals, reflecting the belief that longer range impacts should be discounted at relatively low rates.

The positive discounting literature based on market rates is motivated from infrastructure finance, where capital costs are sourced through borrowing. This literature does not represent regulatory cost incidence very well, given that regulations are financed off-budget through economic adjustments throughout the economy. To capture the extent to which regulations induce economic actors to trade present for future consumption, an economy-wide view is needed. The structural details of credit markets (segmentation and credit constraints) that influence consumer discount rates should be represented in this picture.

The intergenerational discounting literature reflects the point of view that discounting is an appropriate way to address intergenerational tradeoffs. An alternative is to use intergenerational transfers to address intergenerational equity, while using the discount rate to efficiently ration capital in the present.

The implications of regulatory cost incidence for the discount rate and the use of transfer to address intergenerational equity have not received much attention in the literature on environmental BCA. These would be worthy topics for future research.

8.5 Distributional Assessment

Environmental regulations impose uncompensated losses on producers, workers, and consumers. The relative burden of environmental regulations on capital and labor is ambiguous in the literature, as is the impact on sector and economy-wide employment. The empirical characteristics of market structure and the type of regulation affect factors returns and employment levels, and the degree to which producer burdens are passed on to the consumer side of the market.

The literature is relatively inconclusive about the distributional impact of regulations on consumer expenditures. Income-class burdens reflect the policy instrument (tax or performance standard), type of fuel, and institutional factors, such as the indexing of transfer payments to the price level. Heterogeneities also affect household fuel consumption patterns holding income constant. These include household characteristics (e.g., age of residents), local climate conditions, transportation options, and commuting patterns. The price and sources of energy also vary regionally. These factors introduce horizontal distributional differences that are as policy relevant as the vertical distributional impacts of regulations.

The distribution of pollution rents or the use of transfer payments can affect the incidence of regulatory burdens. Also, the capitalization of environmental benefits in property valuations can shift the incidence of local air quality benefits between tenants and landlords.

The full dimensionality of this distributional pattern is hard to represent in RIA, and distributional assessments tend to be relatively limited. Impacts on key constituencies, such as labor in the regulated sector, or small business, are often profiled. This limited recording is likely to miss significant distributional impacts.

CGE model can provide disaggregated surplus measures by income quintiles or deciles that embody all welfare channels. The CGE models can also be linked with bottom-up or PE models to represent sectors or regions in greater detail. The use of CGE models to represent distributional effects may become more common in RIA in the future.

8.6 Uncertainty Evaluation

Strong assumptions about the validity of forecasts are implicit in the use of many applied economic models. Conventional Monte Carlo simulation and traditional sensitivity analyses may not adequately represent forecast uncertainties. Decision-science methods for modeling more fundamental uncertainty are likely to be relevant in many regulatory contexts, especially for regulations addressing climate change. The benefits and costs of expanding the use of fundamental uncertainty evaluation methods in RIA would be a useful area for future research.

In all, the technical challenge involved in the benefit-cost analysis of AEC regulations suggests the need for additional research on the way models are designed, validated, and used to support applied BCA. The uncertainties inherent to the evaluation context also suggest the need for perspective about the information that economic analysis provides for regulatory decision-making. Clarity about the modeling limitations will help decision-makers understand the quality of information that RIA provides.