1.1. Proposed changes in discounting

When OMB first issued guidelines for regulatory analysis, it recommended a default discount rate of 10 %, real and after tax. In 1992, the recommended default rate was lowered from 10 % (real, after-tax) to 7 % (real, after-tax), with additional acknowledgment in the guidelines and related documents (Circular A-94) that lower real after-tax discount rates of 3 and 1 %, respectively, might be justifiable in some cases. These rates were used in 2003, when OMB issued Circular A-4, “Regulatory Impact Analysis.” A distinguishing feature of the OMB approach was to use observed private market interest rates as a (rough) benchmark for the discount rate to be used in benefit–cost analyses. Using a 35 % tax rate to gross-up the OMB-recommended real after-tax rates of 10 and 7 % to real pretax rates of 15 and 11 % and assuming a 3 % inflation rate to convert real pretax rates to nominal pretax rates, yields a market-based benchmark interest rate for discounting of 18 and 14 %, which certainly seemed too high for the twenty-first century.

Moreover, the academic literature on discounting has identified various alternatives using discount rates benchmarked to private market rates. One such alternative, which had been discussed extensively in the academic and policy literature, in the 1980s, 1990s, and 2000s, would be to convert the social costs (in this case of regulation) into foregone annual amounts of private foregone consumption, and then to discount these flows of foregone consumption using individual rates of time preference (approximated by the private return to saving). Another alternative eschews benchmarking the discount rate to private market rates altogether, in favor of using a social discount rate derived not from private market rates, but instead from a normative model of economic growth over time—the so-called Ramsey approach. Yet, another approach shows that uncertainty about the “correct” interest rate can lead to a series of discount rates that decline over time.

The draft revised version of Circular A-4 does a fine job of discussing (1) the general rationale for using a lower real after-tax discount than that recommended in the 2003 version of A-4; and (2) arguments on behalf of using the several alternative approaches mentioned above. At the same time, the particular case for recommending that the new default rate be initially set at 1.7 % can be questioned. Additionally, the draft proposals seemed to be open to different agencies basing their discount rate on one of the alternative approaches.

1.2. Incorporating distributional concerns in regulatory analysis

The draft revisions take up the important question of including equity concerns about the distribution of regulatory benefits and costs in regulatory analysis. An acknowledged feature—some would say limitation—of traditional benefit–cost analysis is its focus on estimating the extent to which a particular government intervention, such as a proposed regulation, is expected to produce aggregate social benefits that equal or exceed aggregate social costs. In aggregating social benefits and costs, the implicit assumption is that $1 of benefit or cost is weighted the same without regard to whom it accrues. The result is that the economic analysis of regulations in regulatory impact analysis fails to provide information about the incidence of regulatory benefits and costs in addition to their aggregate levels. Incorporating distributional effects in regulatory impact analysis requires: (1) estimating how aggregate benefits and aggregate costs are distributed (most likely, although not exclusively, by income) among relevant parties; and (2) developing criteria for “weighting” the different distributional effects. The first requirement is a necessary condition for presenting virtually any type of distributional analysis, but it can be data-intensive. Developing criteria for weighting distributional effects has been discussed extensively in the academic literature, without resolution. The draft revision of A-4 seemed to encourage agencies to consider weighting benefits and costs using a formal social welfare function, giving more (less) weight to benefits received and costs incurred by lower (higher) income groups. My peer review comments observe that simply developing the ability to describe, let alone to weigh, distributional effects, will require developing new analytical capabilities at agencies undertaking regulatory analyses. My peer review comments are more skeptical about the desirability of explicitly weighting benefits and costs using a social welfare function. My main concern is that doing so will produce weighted benefit–cost analyses that blend together both the economic efficiency and distributional equity effects of a regulation into a single bottom-line measure. At a minimum, benefit–cost results should first be presented using unweighted benefits and costs, indicating whether the regulation was economically efficient or not, and then weighted estimates could be presented to show how including distributional considerations changed the results.

Alternatively, regulatory analyses could present (a) a traditional unweighted benefit–cost analysis, and then (b) simply describe, without weighting, the distributional incidence of benefits and costs.

1.3. Concepts of market failure

The 2003 version of Circular A-4 states that regulatory analyses should include a discussion of the market failure(s) that justify regulation in the first place. The proposed revisions retain this guidance, while also broadening possible justifications for regulation to include possible market inefficiencies resulting from decision-making biases of the sort identified and analyzed in the behavioral economics paradigm. The possibility that behavioral decision biases could lead to inefficient outcomes in markets has been recognized, if not completely accepted, in the economics literature, and merits consideration as a rationale for some types of regulation. At the same time, even when decision-making biases may exist, it is not always self-evident that “regulatory experts” are necessarily less prone to their own biases. One implication is that regulatory analyses should be completely transparent about which social benefits of regulation are attributed to traditional forms of market failure, such as externalities, asymmetric information, and so forth, and which social benefits of regulation are attributable to correcting (presumed) decision-making biases. Additional causes of justification for regulation mentioned in the draft A-4 revisions also include limits to competition in markets for goods and services and labor. Lack of competition in goods (and presumably labor) markets could be characterized as a traditional source of market failure. A question, however, is whether these types of market failure are of the sort that are the rationale for the types of regulation—e.g. so-called social as distinct from economic regulation—that are the main focus of OIRA regulatory review.

1.4. Administrative and practical support for regulatory impact analysis

The draft revisions to A-4 provide an excellent overview of ways to improve the quality of regulatory impact analyses. A number of the revisions, such as adopting the social cost of capital approach to discounting or incorporating distributional effects of regulation, will require increased investment in analytical capabilities by agencies, as well as OIRA. An important concern is whether attempts by agencies to adopt OIRA recommendations will lead to a range of regulatory analyses from different agencies that adopt varying approaches to discounting and the incorporation of distributional effects. The ultimate durability and consistency of the regulatory analytic frame will likely require more guidance and input over time from OIRA.

2.1. Peer reviewer

Joseph Cordes, Professor of Economics, Public Policy and Public Administration and International Affairs, Trachtenberg School of Public Policy and Public Administration, The George Washington University; and Co-Director, The George Washington University Regulatory Studies Center.

3.1. Choice of the social discount rate

The proposed revisions to Circular A-4 appropriately devote considerable attention to updating the guidance concerning the discount rate to be used in benefit–cost analysis as applied to government regulations. On balance, the 2023 Circular A-4 covers the main points of the literature on choosing the “correct” social discount rate. The following points should, however, be noted.

1. (1) The specific recommendation to use a default real after-tax discount rate of 1.7 % should be reconsidered on several grounds. First, given the uncertainties involved in choosing the discount rate to pick a single value of 1.7 % introduces an element of false precision. At the very least, an approximation of 2 % (perhaps with some higher alternatives) would seem more defensible. Second, the time frame used to construct the proposed default real social discount rate encompasses a period during which the risk-free Treasury rate was kept intentionally low by quantitative easing and other expansive monetary policy. Thus, while there is a case for proposing a default real discount rate lower than the 2003 Circular A-4 default rate of 7 %, 1.7 % (or even 2 %) may be lower than the longer-run discount rate. Something like a default discount rate of 3 % with a lower value of 2 % and an upper value of 4 or 5 % may be more defensible.

2. (2) The possible role of the Ramsey approach to discounting is unclear. Are agencies being encouraged to use the Ramsey approach as an alternative to the default based on the risk-free real Treasury rate? Moreover, the actual value of the Ramsey rate depends entirely on assumptions made about its components. Depending on these assumptions, it is possible to arrive at a Ramsey rate that is greater than the proposed default rate of 1.7 %. For example, the UK government, which has adopted the Ramsey approach, recommends the use of a 3.5 % real discount rate based on the assumptions of a 1.5 % pure rate of time preference, a value of the marginal consumption elasticity of 1.0, and a long-run growth rate of 2.0 %.

3. (3) The revised draft of Circular A-4 is correct in noting that when capital market distortions, such as taxes, drive a wedge between the real return to capital and the individual discount rate, the conceptually correct approach for discounting is the so-called shadow price of capital approach. The challenge is the practical and consistent implementation of this method. The revised Circular A-4 presents an example based on the Resources for the Future (RFF) Working Paper by Newell, Brest, and Pizer, that could serve as a template by agencies. To the extent that OIRA wishes to encourage agencies to consider using the shadow-price approach, it would be useful to provide somewhat more detailed examples of how to implement the shadow price approach, if not in the circular itself, then in a web link. An approach consistent with the shadow price of capital framework can also be found in Mannix (OMB-2022-0014-3906).

4. (4) Finally, the proposed revision of Circular A-4 presents the case for possibly applying a schedule of declining discount rates for policies (mainly environmental) with intergenerational benefits. There is some support for this in the peer-reviewed literature cited in the revised Circular A-4. A concern, however, is that the precise level and “shape” of the declining discount rate schedule is likely to be quite sensitive to specific assumptions made about the underlying distribution of “uncertain” discount rates. Indeed, the specific schedule of declining discount rates that is presented in the preamble is offered with relatively little justification.

3.2. Distributional effects in regulatory impact analysis

Compared with the 2003 Version of Circular A-4, the proposed 2023 revision devotes considerable attention to incorporating distributional effects of regulation into regulatory impact analysis. Identifying and, where possible, estimating the distributional effects of proposed government regulations is clearly desirable. The question is how best to do so.

The critical first step in any analysis of distributional effects is a determination of the incidence of aggregate benefit and aggregate costs of regulation. Namely, how are both the projected benefits and the projected costs of regulation distributed among the relevant groups affected by the regulation? The word “relevant” is underscored because although a typical assumption may be that what matters is distribution by income group, this need not necessarily be the case, depending on the regulation under consideration. For example, in some cases, the relevant distributional effects may be geographical. Agencies would benefit from more guidance on this matter. Such guidance need not be incorporated in Circular A-4 itself, but perhaps could be provided in a link to a supplementary website focusing on “how to incorporate distributional effects.”

There are formidable conceptual and empirical challenges to identifying and estimating distributional effects. Simply determining who benefits and who bears the cost requires determining what public finance economists refer to as the economic incidence of regulatory benefits and regulatory costs. This may or may not correspond to what might be described as the initial impact of the regulation. Consider the case of a regulation that requires producers to reduce carbon emissions. The costs of complying with such regulations may be paid by producers, but may be shifted backward to workers through lower wages and forward to consumers in the form of higher prices.

Additional guidance to agencies on the basics of incidence analysis may be useful, indeed required, to facilitate some consistency in how distributional effects are analyzed. A possible model may be to adapt the manner in which economic incidence analysis of taxes is undertaken by the United States Treasury and the Congressional Budget Office to determine the economic incidence of regulation. Indeed, it may be helpful to develop guidelines for undertaking distributional transfer analysis, analogous to benefit transfer analysis often used in environmental benefit–cost analysis.

Once the incidence of regulatory benefits and costs has been identified and estimated, the issue is that of how best to incorporate distribution into regulatory impact analysis. Although the proposed Circular A-4 revision recognizes that there are various ways of taking distribution into account, its quasi-endorsement of “distributionally weighted” benefit–cost analysis is problematic in several ways.

1. (1) The acknowledged purpose of the benefit–cost analysis is to assess whether a particular policy change, such as a government regulation, enhances economic efficiency. While, from a welfare-analytic perspective, government regulations should be both economically efficient and have desirable (or at least acceptable) distributional outcomes, the two objectives are distinct. The public interest is best served by presenting separate impacts of regulation on efficiency and equity rather than by attempting to combine the effects in a single weighted benefit–cost measure.

2. (2) The conceptual premise behind the suggested method of constructing distributional weights is relatively weak. As several public commenters—see Banzaf (OMB-2022-0014-0158), Sullivan (OMB-2022-0014-0029), Kenkel (OMB-2022-0014-3910), and Fraas (OMB-2022-0014-3917)—have noted, the common argument that it is reasonable to assume that the marginal utility of an additional $! of income may decline with income for a single individual is based more on subjective views of what seems reasonable, rather than on empirical evidence; moreover, it does not necessarily apply across individuals.

The preferable approach would be to incorporate distributional effects in a regulatory impact analysis in much the same way as distributional effects are presented separately from efficiency effects in the analysis of tax policy. Namely, show how the benefits and costs are distributed among the relevant groups in addition to and separately from presenting any estimates of the policy’s impact on economic efficiency. The inherently subjective weighting of these separate effects is best left to decision-makers and the political process.

3.3. Geographical scope

Guidance in the 2003 version of Circular A-4 states that the “….analysis should focus on benefits and costs that accrue to citizens and residents of the United States.” In the case of a regulation that “is likely to have effects beyond the borders of the United States, these effects should be reported separately.” Thus, while the inclusion of benefits and costs beyond the borders of the United States in regulatory impact analyses was not expressly prohibited in the 2003 version of Circular A-4, the inclusion of benefits and costs experienced by non-US citizens is circumscribed.

The greater willingness to consider global benefits and costs that is found in the 2023 proposed revisions to Circular A-4 reflects a legitimate debate about how to treat “global social benefits and/or costs,” particularly in the case of environmental problems whose scope is global rather than national—for example, Smith (OMB-2022-0014-0079). One way of framing the issue in a manner that is entirely consistent with established benefit–cost principles is as follows. If one accepts the premise that the relevant social benefits should be based on the willingness to pay for environmental improvement of US citizens, the question can be reframed as follows: (1) To what extent do US citizens have a positive willingness to pay for environmental benefits that accrue to citizens in other countries, and if so, (2) how should this willingness to pay be estimated?

Scholarly research suggests that the answer to the first question is “yes,” while the answer to the second question may be that $1 of environmental benefit accruing to citizens of other countries would be valued at <$1 of benefit accruing to US citizens. This suggests that a conservative approach to incorporating global benefits and costs would be: (a) to include global benefits and costs separately, along with purely domestic benefits and costs, as recommended in the 2003 version Circular A-4, and (b) present a range of values for such global benefits, treating the full magnitude of global benefits and costs as “upper bound” estimates, and applying an appropriate discount to such values to represent the willingness to pay of American citizens. The revised draft is generally consistent with this approach; however, the brief reference to situations where “such effects cannot be separated in a practical and reasonably accurate manner, or that the separate presentation of such effects would likely be misleading or confusing in light of the factors detailed above” is likely to confuse agencies and lead to inconsistencies across analyses.

3.4. Incorporating insights from behavioral economics

The revised circular includes a discussion of the possible role of behavioral economics in regulatory impact analysis. The suggestion to draw upon behavioral economic insights, such as through the use of nudges to implement regulations, is useful. A potential caveat is that a recent survey of empirical studies (Mertens et al. (2021), “The effectiveness of nudging: A meta-analysis of choice architecture interventions across behavioral domains,” Proceedings of the National Academy of Sciences) finds that nudges have small to modest effects on outcomes.

Caution, however, should be exercised in broadening the list of possible justifications for regulation to include “behavioral biases.” As Geyer and Viscusi (Journal of Benefit Cost Analysis, 2016) note:

To the extent that behavioral biases are offered as a justification for government intervention, and are used as the basis for estimating presumed social benefits of intervention, agencies should be strongly encouraged to adopt something like the Geyer–Viscusi “behavioral transfer” test.

Moreover, presentation of estimated benefits in a regulatory impact analysis should separate those benefits arising from correcting “traditional” market failures from those that are attributable to behavioral biases.

3.5. Discussion of uncertainty

Proposed revisions devote more attention to uncertainty than does the 2003 version. At times, however, the discussion of uncertainty seems to conflate two distinct aspects of when uncertainty is relevant. One important issue is how to best account for the unavoidable uncertainties encountered in estimating regulatory impacts. The revisions do provide a brief discussion of what is commonly called sensitivity analysis. At a minimum, undertaking even a simple incremental sensitivity analysis should be considered a strongly recommended best practice in regulatory impact analysis. More sophisticated forms of sensitive analysis based on Monte Carlo simulations are increasingly accessible in Excel-based programs such as Crystal Ball, and agencies should be encouraged to adopt these technologies.

The other dimension of uncertainty discussed in the revision pertains to what assumptions are appropriate about how individuals and businesses incorporate uncertainty into their decisions. It is not always clear, however, about how different assumptions about uncertainty—for example, risk neutrality versus risk aversion—should be incorporated in regulatory impact analyses.