RESEARCH DOSSIER

Intergenerational Choice and Social Discounting

Abstract. Intergenerational choice concerns the allocation of resources, welfare, risks, and obligations among persons living at different dates. This monograph studies the conceptual, mathematical and institutional foundations of social discounting, distinguishing pure utility discounting, marginal consumption valuation, social opportunity cost of capital, state-dependent valuation and public appraisal conventions. It reconstructs the deterministic Ramsey rule, derives stochastic marginal-utility and uncertain-rate results under explicit assumptions, examines cohort incidence, population aggregation, nonmarket relative prices and the limits of infinite-horizon comparisons, and analyzes the consequences of alternative term structures for long-lived public investments. Documentary comparisons distinguish the October 2026 United Kingdom Green Book, the United States regulatory convention reinstated in 2025, and the Norwegian expert committee's historical recommendation of 2012. Independently calculated numerical illustrations make rate sensitivity and forward-rate accumulation reproducible. The original analytical synthesis identifies how welfare premises, capital displacement, distributional incidence and administrative rules produce distinct rankings. The research is a theoretical and documentary synthesis, with source-specific access limitations and hypothetical numerical examples explicitly distinguished from empirical estimates.

Keywords: Intergenerational choice; social discounting; social welfare; Ramsey rule; pure time preference; consumption discount rate; stochastic discount factor; declining discount rates; public investment; cost-benefit analysis; capital displacement; future generations; distributional incidence

The economic object and the units of comparison

Intergenerational choice is the selection of feasible economic paths that distribute benefits, costs, risks, resources, and institutional obligations among persons living at different dates. Social discounting supplies the intertemporal conversion factors used in such evaluations. A discount factor is a present-value multiplier for a specified future unit; a discount rate is a parameter or a derivative of the corresponding factor. The unit is indispensable: an additional unit of consumption, a unit of utility, a financial payment, an avoided illness, and an intact ecosystem are different objects. A conversion factor for one cannot be transferred to another without a theory of valuation.

Consider a reservoir that costs resources now, reduces flood exposure for many decades, and creates maintenance liabilities. The contractor's return, the treasury's cash account, the expected household losses avoided, and the welfare of households that have yet to be born answer different questions. The social account includes resource displacement, environmental effects, and incidence. The financial account includes payments and receipts. Where payments are transfers within the defined population, their distribution matters although their gross amount is not itself a net resource loss. The correct accounting boundary must precede the discount calculation.

The word generation can describe a birth cohort, persons alive in a given period, or a succession of representative decision makers. A period-based utility sum counts utility flows, whereas a cohort-based criterion first aggregates the consequences for each birth cohort across its lifetime. These two operations can produce different rankings where lifetime length, migration, survival, or the number of persons changes. In the present study, the mathematical benchmark is a period-based consumption welfare function with an explicitly fixed population unless otherwise stated. Cohort aggregation enters separately. The units of account, the baseline, the affected population, and the counterfactual supply the common foundation for mathematics and institutional comparisons.

Goulder and Williams (2012) distinguish welfare and finance-equivalent rates. This distinction is the organizing issue of the dossier: a number appropriate to a particular accounting convention carries no automatic authority in another convention. The inquiry proceeds through formal benchmarks, comparative institutions, and explicit examples. It is a theoretical and documentary synthesis rather than an econometric identification of an optimal universal rate.

Historical foundations of saving and welfare

Ramsey's 1928 mathematical theory of saving established the historical problem of how an economy should allocate consumption across time. An efficient path of accumulation has opportunity costs: additional capital requires foregoing resources that could have been consumed. It also changes the feasible set of subsequent generations. The decision concerns an entire path, rather than a single isolated exchange. As the historical account by Attanasio (2015) explains in his reassessment of Ramsey, the contribution anticipates important developments in modern growth and intertemporal economics.

The modern Ramsey rule should be read as a conditional result from a particular utility and growth specification. Its use in project evaluation is a local welfare comparison, while the original saving problem concerns the choice of a feasible aggregate path. Confusing the two permits an analyst to treat the projected growth rate as simultaneously an exogenous forecast and the optimal outcome of the policy being evaluated. For marginal projects, the reference path can be held fixed; for transformational policies, the growth path itself is an outcome.

The second historical development is the discounted utility representation of preferences. Its mathematical usefulness stems from additive separability across dates and the ability to assign stable weights. Private time preference, however, reflects individual horizons, uncertainty over access to resources, borrowing conditions, liquidity, and anticipated experiences. Frederick, Loewenstein, and O'Donoghue (2002) examine these issues in their review of time preference. The behavior of living persons does not by itself reveal the ethical standing of unborn persons.

The third development is the conversion of welfare principles into administrable public appraisal. Standard schedules allow agencies to compare proposals on a common basis, protect against strategically selected parameters, and make budget decisions reproducible. Their institutional function is significant even when the chosen number is imperfect. Economic science supplies the comparative logic and conditions for these conventions. Political and legal institutions determine the legitimate decision procedure. The rest of this monograph keeps both aspects visible.

Present value arithmetic and term structure

For deterministic annual real consumption-equivalent net benefits b_t, let D_0=1 and D_t=1/(1+r)^t for a constant effective annual rate r greater than −1. The net present value is

NPV=∑t=0TDtbt.NPV=\sum_{t=0}^{T}D_t b_t.

The benefit at date zero is already in the numeraire. For a sequence of effective annual forward rates f_j, the correct factor is

Dt=∏j=1t(1+fj)−1.D_t=\prod_{j=1}^{t}(1+f_j)^{-1}.

The equivalent annual spot rate is D_t^{-1/t}−1 for t>0. Thus a lower prescribed rate beginning in year 31 applies to the incremental years, not retroactively to the first thirty. A curve with 3% for thirty years and 2.5% thereafter does not assign a year-40 factor equal to 1/(1.025)^40. It assigns 1/(1.03)^30/(1.025)^10.

For continuously compounded rates, D(t)=exp(−∫_0^t f(s)ds). The average rate is −log D(t)/t and the instantaneous forward rate is −d log D(t)/dt. They agree for a flat curve and generally differ for a nonflat one. Converting annual discrete rates to continuous rates requires log(1+r); adding an effective annual growth percentage to a continuous rate without such conversion is only an approximation.

Inflation requires an independent transformation. If a nominal amount B_t is expressed relative to a general price index P_t with base P_0, its real value is B_t P_0/P_t. With deterministic annual inflation π and a matching nominal interest rate i, the identity is 1+i=(1+r)(1+π). Deflating a nominal flow and discounting at a nominal rate counts inflation twice. Real relative prices of particular goods may still change after general inflation has been removed. HM Treasury's Green Book discounting guidance (2026) emphasizes the separation between the general price adjustment and the real appraisal rate.

These identities also explain why reporting only a headline rate is insufficient. A reproducible appraisal records the compounding convention, timing of flows, base-year price level, full term structure and treatment of terminal effects. An apparent disagreement between two analysts may be a disagreement about conversion conventions before it is a disagreement about welfare.

The Ramsey welfare benchmark

Assume deterministic consumption c(t)>0 per person, fixed population, time-additive welfare and a differentiable, increasing, concave utility function u. Consider

W=∫0Te−ρtu(c(t))dt.W=\int_0^T e^{-\rho t}u(c(t))dt.

Here ρ is the pure utility discount rate. For infinitesimal consumption changes, the ratio of marginal social value of a unit at time t to a unit now is D_c(t)=exp(−ρt)u'(c(t))/u'(c(0)). Under constant relative risk aversion utility, u'(c)=c^{-η}; for η=1 the utility function is logarithmic, while for η≠1 it can be written (c^{1−η}−1)/(1−η). Differentiation yields

rc(t)=ρ+ηg(t),g(t)=dlog⁡c(t)dt.r_c(t)=\rho+\eta g(t),\qquad g(t)=\frac{d\log c(t)}{dt}.

This is the continuous-time marginal consumption discount rate. With annual growth g_d and an annual pure utility rate δ, the exact discrete factor per year is (1+δ)^{-1}(1+g_d)^{-η}, and the annual consumption rate is (1+δ)(1+g_d)^η−1. The distinction matters where quantities are high or the horizon is long.

Positive growth lowers the relative marginal utility of an additional unit for a future person if utility is concave. A zero pure utility rate is therefore consistent with positive consumption discounting. Conversely, negative growth can imply a negative consumption discount rate, even with positive pure time preference. The value of η measures curvature in this representation: it determines how strongly relative marginal utility changes with consumption. Its interpretation as risk aversion, contemporaneous inequality aversion and intertemporal substitution coincides only under particular separable preference assumptions.

Arrow and colleagues (2014) connect Ramsey discounting and project appraisal; the expert panel reported by Arrow and colleagues (2012) similarly treats the Ramsey organizing framework as useful for understanding long-horizon decisions. The equations above independently reconstruct this benchmark. They are not an empirical claim that growth is constant or that society embraces a particular curvature.

A finite project can alter the consumption trajectory materially. The first-order approximation then may omit second-order changes in marginal valuation and interactions with capital accumulation. For such decisions evaluate W on each feasible path using an explicitly specified welfare objective, or report an interval and alternative objectives. A discount rate inferred on the baseline is a useful local shadow price, not necessarily the correct global transformation for every counterfactual.

Pure time preference and ethical authority

The pure utility discount rate is a normative intergenerational weight in a social welfare representation. If ρ is positive and everything else equal, the utility increment of an otherwise identical person receives a lower marginal weight solely because it occurs later. When the object is within-person allocation, similar mathematics can represent impatience. In a collective decision spanning birth cohorts, the justification instead concerns the moral and institutional status assigned to people at different historical times.

A survival adjustment requires its own interpretation. Suppose an extinction event with constant independent hazard h removes the possibility of future utility under every policy. Conditional welfare is multiplied by exp(−ht), which behaves algebraically like an additional utility discount rate in this narrowly defined setting. If survival probability is policy-dependent, however, it is part of the outcome being evaluated. Reassigning it to a common discount rate removes the channel through which prevention affects survival. A loss of benefits because a bridge fails likewise belongs in expected project consequences rather than a universal ethical parameter.

Private experiments commonly expose people to delayed payments within finite lives. They can inform a model of behavior, but their estimates also incorporate credibility, liquidity and transaction constraints. The distinction follows from the subject matter reviewed by Frederick and colleagues (2002) in discounting and time preference. A representative individual's desire for immediate payment does not settle the constitutional question of intergenerational representation.

HM Treasury (2026) makes a particularly explicit institutional distinction in its response on pure time preference: the retained 0.5% parameter is intended as social pure time preference, with mortality and catastrophic risks dealt with separately. This is evidence of a specified public convention, not a demonstration that 0.5% is ethically correct universally.

Normative analysis must state whether equal standing is understood as equal marginal utility weights, equal treatment of a fixed consumption unit, equal opportunity, a minimum living standard, or equal representation of claims. These interpretations can diverge. A society can assign equal utility weight to persons and still favor a transfer to poorer persons, whichever period they inhabit. Conversely, equal monetary weights can neglect inequality if future persons have very different marginal utilities. Explicit ethical premises make disagreements inspectable and preserve the distinction between assumptions and results.

Growth, inequality, and the distribution of consumption

Aggregate growth is a statement about the total or average measured quantity. It does not guarantee that a given person or location receives the average. Let c_{it} denote consumption of group i at date t and λ_{it} a welfare weight. The marginal present valuation of a change in that group's consumption is proportional to exp(−ρt) λ_{it}u'(c_{it}). Aggregate appraisal using a single consumption discount factor therefore requires either a representative-agent abstraction, a preexisting adjustment to the distribution of effects, or special restrictions on marginal utilities.

For illustrative purposes, let two equally weighted households have consumption 1 and 4 in the base year. With η=1 the respective marginal utilities are 1 and 0.25. The same small unit directed to the first household has four times the marginal utility effect. A future policy that raises average consumption but imposes uncompensated costs on the poorer group may have positive aggregate consumption NPV and negative distribution-weighted welfare. The loss is invisible if benefit and cost streams are aggregated before incidence is recorded.

The distinction between rates and weights is especially important in global public goods. A country-level schedule may describe how a national treasury evaluates the opportunities of its residents. Applying this schedule to uncompensated damages abroad embeds a boundary decision about whose welfare counts. A global judgment requires a defined international incidence account and purchasing-power or consumption conversion consistent with the welfare criterion. The fact that two monetary amounts share a currency after exchange-rate conversion does not make their welfare significance equal.

In an economy where average consumption grows at 2% annually while a vulnerable cohort's consumption remains flat, the deterministic representative Ramsey formula with η=1 and ρ=0 yields approximately 2% for the aggregate path and 0% for the vulnerable cohort, prior to base-level marginal-utility normalization. This is a constructed comparison, not a prediction of the world's income distribution. Project appraisal is strongest when it exposes these differences rather than treats growth as an unconditional warrant for lower concern about the future.

Drupp and colleagues (2018) identify the importance of disentangling social discount-rate components. The lesson for distributional analysis is that a headline rate does not disclose the growth, curvature and ethical assumptions behind it.

Risk, uncertainty, and stochastic discount factors

Assume expected time-additive CRRA welfare with fixed population, initial consumption known, and a stochastic future consumption path. The marginal stochastic discount factor for a unit delivered at time t in state ω is

Mt(ω)=e−ρt(ct(ω)c0)−η.M_t(\omega)=e^{-\rho t}\left(\frac{c_t(\omega)}{c_0}\right)^{-\eta}.

A marginal uncertain net benefit B_t has present consumption-equivalent value E[M_t B_t]. Writing this as E[M_t] E[B_t] + Cov(M_t,B_t) reveals two different uncertainties: uncertainty about the value of consumption and uncertainty about the project's delivery. A project that yields more benefits in states with low consumption has positive covariance with M_t under this model, making it relatively valuable as insurance. A project whose benefits arrive mainly in prosperous states has a different valuation.

For a riskless consumption unit, D(t)=E[M_t]. If log(c_t/c_0) is normally distributed with mean μ_t and variance v_t, then

−log⁡D(t)t=ρ+ημtt−η22vtt.-\frac{\log D(t)}{t}=\rho+\eta\frac{\mu_t}{t}-\frac{\eta^2}{2}\frac{v_t}{t}.

This result is exact under the stated lognormal assumption; μ_t is expected log growth over the horizon rather than expected arithmetic growth. If log growth increments are independent with fixed drift and variance, both μ_t/t and v_t/t are constant. The spot rate is then constant even though uncertainty exists. Declining rates require additional features such as persistence, parameter uncertainty or changes in the distribution of cumulative growth.

The expected marginal-utility formula illustrates why a risk premium is not a universal addition to every public cost. If a project prevents resource losses during recessions, a conventional positive benefit-risk premium may penalize precisely the states where its benefits are socially valuable. If future resource costs increase during shortages, their covariance can raise the present cost. A single project-wide risk premium loses this asymmetry. The proper approach is scenario-by-scenario net valuation when credible joint probabilities and impacts are available, with separate sensitivity analysis where they are not.

Arrow and colleagues (2014) discuss correlated consumption growth shocks as one basis for declining certainty-equivalent rates. The derivation here clarifies its conditional structure: uncertainty influences future valuation through the distribution of marginal value, and its maturity pattern depends on the stochastic process.

Uncertain rates and the arithmetic of declining schedules

Consider an illustrative model in which a continuous discount rate R is drawn once and remains constant throughout the horizon. The decision is made before the draw, and a fixed future consumption-equivalent benefit has present factor D(t)=E[exp(−Rt)]. Assume finite exponential moments and sufficient differentiability. Define tilted expectation E_t[X]=E[X exp(−Rt)]/D(t). Then

f(t)=−dlog⁡D(t)dt=Et[R],f′(t)=−Var⁡t(R)≤0.f(t)=-\frac{d\log D(t)}{dt}=E_t[R],\qquad f'(t)=-\operatorname{Var}_t(R)\leq 0.

The forward rate declines because low-rate states contribute increasing shares of present valuation. This is a result about a particular probabilistic model. It does not state that every future spot interest rate falls over time or that the project discount factor is independent of benefits. For a finite discrete distribution with a strictly lowest rate assigned positive probability, the tilted expectation converges to that rate. If all rates coincide, the variance is zero and there is no decline.

For a gamma-distributed nonnegative rate with shape k and scale θ, the Laplace transform gives D(t)=(1+θt)^{-k}; hence f(t)=kθ/(1+θt). This is a convenient analytical illustration of Weitzman's (2001) gamma discounting. It produces a hyperbolic-type factor from heterogeneous or uncertain exponential rates, rather than from an assumed individual's hyperbolic impatience.

Gollier and Weitzman (2010) examine uncertain distant-future rates in a decision problem designed to resolve an apparent inconsistency between discounting and compounding arguments. Their model clarifies why the valuation perspective matters. In our reconstruction, the random rate is fixed across time; changing to an independently drawn new rate each period changes the result. With independent, identically distributed one-period discount factors q_j, E[∏q_j]=(E[q])^t, yielding a constant equivalent annual rate. Neither case establishes the true stochastic structure of real economies for a century.

The broader review by Arrow and colleagues (2014) concludes that the case for declining certainty-equivalent schedules merits institutional consideration. In application the persistence assumption, state dependence, and the distinction between rate uncertainty and ethical disagreement must remain explicit. Averaging rates is not equivalent to averaging discount factors; because exp(−Rt) is convex in R for t>0, the expected factor is at least the factor at the expected rate. An arithmetic average of proposed percentages is thus not a neutral aggregation rule for future values.

Two illustrations of uncertainty and horizon sensitivity

The following calculations use deliberately hypothetical inputs and are reproducible from the displayed equations. There is no fitted statistical model or claim about actual project returns. First, let the continuous rates 1% and 5% each have probability one half and persist for the entire horizon.

D(t)=0.5e−0.01t+0.5e−0.05t.D(t)=0.5e^{-0.01t}+0.5e^{-0.05t}.

In the same model the instantaneous rate is the valuation-weighted mean of 1% and 5%. Its initial level is 3%; it subsequently moves toward 1%. Contrast this with a constant continuous 3% curve. The arithmetic mean rate is the same at the initial date; the future factors are different. At long maturities the comparison becomes increasingly important for a fixed future benefit, because high-rate scenarios have little present-value weight. Neither curve includes the covariance of project delivery with consumption.

Second, consider a public asset with an immediate cost of 100 units and a single benefit of 400 after 50 years, measured in constant-price consumption. For a constant effective annual rate r, NPV=−100+400/(1+r)^50. The break-even rate is 4^(1/50)−1, approximately 2.81% annually. A higher rate rejects the project under monetary NPV, a lower rate accepts it. If the benefit is an environmental improvement whose real relative value rises before delivery, its future monetary amount must be adjusted separately; that is a change in the payoff, not evidence that the consumption discount rate itself has changed.

A ranking reversal between this project and an immediate alternative can identify which assumptions determine disagreement. It cannot select the ethically optimal parameter by itself. Reporting a threshold is analytically more informative than presenting one positive or negative NPV. For streams that change sign multiple times, several internal rates may exist or none may be economically useful. The entire NPV profile and feasible alternatives should then be examined.

The empirical literature adds evidence about persistence, rather than eliminating the need for uncertainty analysis. Newell and Pizer (2003) use historical US interest-rate data and compare alternative stochastic models. The paper shows why the distant future becomes sensitive to assumptions about the behavior of rates. The examples here do not replicate its estimates, but isolate the underlying arithmetic so that it can be evaluated independently.

Opportunity cost, capital displacement, and the financial account

The social opportunity cost of capital is the value of consumption and investment opportunities displaced by a public project. The social rate of time preference, by contrast, expresses an intertemporal marginal trade-off in consumption under a specified welfare or behavioral account. In an idealized frictionless competitive economy with aligned marginal conditions, a suitable market return and the consumption discount rate can coincide. Taxes, risk, externalities, credit restrictions, and incomplete markets disturb that equality.

A shadow-price-of-capital approach converts displaced investment into consumption-equivalent streams or values before applying a consumption discount factor. Consider a unit of resources financed from consumption and a unit financed by reducing a private investment that would have generated future net output. The two units have different opportunity costs even though the budget expenditure is identical. If the displaced investment's future consumption-equivalent value is included explicitly, applying an additional capital-return premium to the same displacement can double count its effect.

Goulder and Williams (2012) distinguish a potential Pareto improvement test from a welfare test under an existing social welfare function. If a policy could in principle compensate its losers, a surplus test may identify an efficiency opportunity. Actual compensation, however, requires institutions, fiscal capacity, and feasible transfers across time. A person who is not yet born cannot sign a complete contingent contract today. A positive potential surplus does not establish that the specified distribution of policy outcomes is ethically preferred.

The market yield on a long-term government bond often provides a useful reference to financial opportunity costs. Its magnitude can nevertheless reflect inflation expectations, liquidity, default risk, term premiums, taxation and monetary-policy expectations. A nominal corporate hurdle rate is a different observation. Equity returns compensate systematic exposure in an asset market and cannot be assigned mechanically to riskless consumption benefits or to ecological damages.

An economically coherent appraisal therefore reports both the resource funding account and the consumption welfare account when they differ substantially. It explicitly identifies capital displacement and the assumed attainable rates of return. A public project that yields social protection in bad states may have a low financial return and substantial welfare value. A project offering high financial revenue may impose external costs. Independent accounts preserve the information needed to compare these situations.

Capital accumulation and overlapping generations

A representative infinitely lived household model compresses many distinct ages and future persons into an intertemporal choice problem. Overlapping-generations analysis exposes a different structure. Suppose each cohort works when young and consumes savings or transfers when old. With no bequest, the welfare of an unborn cohort is not part of current household utility unless an institutional or altruistic connection is introduced. A pension system, public debt instrument, or long-lived infrastructure stock can carry consequences across this boundary.

A stylized competitive economy with constant population growth n and no technological growth can have a steady-state capital stock per worker k. Let f(k) be output per worker and d depreciation. Steady-state consumption per worker is f(k)−(n+d)k in the usual continuous accounting approximation. An interior maximum of steady-state consumption satisfies f'(k)=n+d. If the economy has accumulated beyond this golden-rule stock, a reduction in accumulation can increase feasible steady-state consumption. This result concerns a particular feasible production model and should not be confused with a theorem that all observed capital markets are inefficient.

The economic interpretation of public debt depends on the cohort distribution of both liabilities and assets. A domestic government bond is a liability of taxpayers and an asset held by someone. Future debt service may redistribute within a cohort or between cohorts; real effects depend on displaced capital, distortions, refinancing risk and distribution. Gross debt alone does not measure the intergenerational burden. At the same time an accounting offset between assets and liabilities does not remove real resource consequences.

A public investment can combine present borrowing, future taxation, and durable public benefits. If a cohort receives the services while another pays, all three channels must be identified: resource costs, financial obligations, and benefit incidence. The analyst should compare the policy with a counterfactual financing and investment path. Assuming a single social rate before the financing mechanism is defined suppresses an important part of the causal question.

Attanasio's (2015) historical link to growth theory places Ramsey's work in the context of later dynamic models, including the problem of agents with different birth dates. The illustrations in this section are independent benchmark calculations. The analysis establishes why capital efficiency and intergenerational distribution belong to a common study while remaining different economic questions.

Population, cohort welfare, and infinite horizons

Population changes make the object of intergenerational welfare especially consequential. Let N_t be population, c_t per capita consumption and u(c_t) utility. A total period utility criterion takes the form Σ β^t N_t u(c_t). A per capita period criterion takes the form Σ β^t u(c_t), and a cohort criterion may assign weights to lifetime well-being instead. A constant added to utility changes comparisons under total utility when populations differ, even though it leaves fixed-population comparisons unchanged. Thus choice of normalization and aggregation cannot be treated as cosmetic.

Population-dependent uncertainty also influences the treatment of catastrophic risk. A policy affecting future survival or the number of persons changes the domain of comparison. A marginal consumption formula conditional on a fixed surviving population need not capture these consequences. Assigning a price to a reduction in a current person's mortality risk differs from resolving a choice over the existence and identity of future people. Economic appraisal can expose these differences without presuming that all population ethics debates have settled answers.

Infinite horizons require mathematical care. If c_t is constant, utility is strictly positive and ρ=0, an undiscounted infinite sum diverges. Comparing two paths through their raw infinite totals then fails. An additive constant in u may itself create divergence. Finite-horizon comparisons, overtaking criteria, and sustainable preference orderings are different ways of posing well-defined intergenerational questions; each entails restrictions on what comparisons are possible.

Chichilnisky's (1996) axiomatic sustainable development examines criteria designed to prevent exclusive domination by either end of the temporal horizon. This line of inquiry makes visible the ethical choice inherent in the aggregation rule. A discounted utility sum on a bounded stream has a diminishing sensitivity to a fixed tail change when the discount factor is summable. Yet a criterion sensitive only to limiting outcomes could overlook severe transitional suffering. The issue is the structure of social preference over sequences.

A finite appraisal horizon is an operational choice, not a substantive assertion that persons beyond the horizon have no claims. Where effects persist, terminal valuation, physical feasibility conditions, or separately reported continuation scenarios are necessary. A project whose environmental damage falls beyond the formal cutoff should not be described as harmless simply because the monetized account ends earlier.

Nonmarket goods and changing relative prices

The fundamental numeraire problem returns in environmental, health and cultural appraisal. Let q_t denote an environmental service and p_t its marginal consumption-equivalent value in real units. For a small change Δq_t, b_t=p_t Δq_t. If p_t grows at a continuous rate h under a specific model, and consumption benefits are discounted at continuous rate r, the present value of a constant expected physical effect at date t scales as exp[−(r−h)t]. The effective decay of the monetary value of that physical effect differs from the discount rate on consumption.

The distinction is essential for scarce ecosystems: ecological goods can become more valuable relative to ordinary consumption as scarcity increases. It is also essential for health gains if their consumption-equivalent valuation changes with income or health technology. The quantity and shadow-price paths are empirical or model-based objects, with uncertainty requiring independent specification. A blanket sector-specific lower rate can approximate a particular relative price growth path, but it can conceal the assumptions and accidentally count growth in value twice.

The 2026 UK response on relative price adjustments explicitly supports using projected relative valuations for non-consumption outcomes with a common consumption discounting framework. The institutional rule provides a direct example of this separation. It does not eliminate the difficulty of estimating the value of an endangered species, avoided long-term morbidity or irreversible ecosystem loss.

A marginal willingness-to-pay measure also reflects the affected person's ability to pay. If a low-income population experiences severe risks, a low estimated monetary payment capacity is not sufficient evidence of low welfare loss. Distributional adjustment, alternative outcome measures, or explicitly stated social objectives may be needed. In a constrained ecological decision, physical minimum standards may define admissible policies before monetary rankings are calculated.

Some consequences cannot be credibly monetized at the necessary horizon. Their omission from a monetary total is a measurement limit, not a zero valuation. An appraisal can present the monetary NPV alongside quantified physical outcomes, intervals, and clearly justified qualitative constraints. Doing so preserves the decision relevance of effects whose shadow prices are poorly identified.

Climate policy as a test of intergenerational reasoning

Climate policy combines a long-lived stock externality with uncertain economic and ecological consequences. Current mitigation changes the future trajectory of damages; adaptation alters vulnerability; outcomes vary by region and income. Stern's (2007) chapter on economics and climate ethics identifies the global and persistent nature of this problem and the possibility of large changes. A single future cash flow is therefore a poor representation of the entire economic object.

Nordhaus (2007), in his review of the Stern approach, argues that its strong policy conclusions are highly sensitive to near-zero utility discounting and particular utility assumptions. The critique links normative choices to observed economic behavior and market-consistent calibration. The comparison is scientifically valuable because it exposes the dependence of recommendations on parameter selection rather than merely contrasting preferred policy outcomes.

A model of climate policy requires physical emissions and concentrations, a damage specification, technological transition costs, growth, adaptation, and a welfare criterion. For a marginal emissions pulse, the consumption-equivalent social cost is the discounted expected value of incremental damages along a defined baseline. In a stochastic model it is appropriate to value damages with the same marginal-utility state dependence derived earlier. Damage estimates and the discount factor are jointly relevant when climate outcomes influence consumption.

The effect of lower discounting on policy generally increases the weight on delayed consequences, but not all policy consequences have the same sign. Future decommissioning costs, maintenance burdens, infrastructure lock-in, and induced damages also gain importance. A project cannot become desirable by labeling all future effects as benefits. If benefits and costs both occur in the distant future, relative timing and covariance determine the effect on their net present value.

The substantive conclusion of comparing the Stern and Nordhaus positions is thus conditional. Their disagreement concerns welfare specification, calibration, time horizon, model structure, and treatment of uncertainty. A model comparison should hold the damage model fixed while varying discount parameters, then hold discounting fixed while varying damage functions, and finally examine interactions. Without that decomposition, attribution of a policy ranking to the discount rate alone remains incomplete.

Long-lived infrastructure, prevention, and irreversibility

Long-lived assets involve initial construction, repair, replacement, operations, and eventual disposal. Avoidance projects such as flood defenses, pandemic readiness, and pollution prevention have different state-dependent benefits from a facility producing saleable output. Their value depends on the probability and severity of events with and without the policy. The project-specific joint distribution is as important as the maturity of the expected payoff.

A deterministic benchmark might approximate a perpetual constant real benefit b occurring at each year-end with D_t=(1+r)^{-t}. For r>0 the present value is b/r; if the stream begins only after year m, its present value at zero is (b/r)/(1+r)^m. If a real benefit grows at a constant effective annual rate g<r, the value for first-year benefit b_1 is b_1/(r−g). These formulas require a continuing feasible stream, a constant compatible rate, and convergence. They are not appropriate for an asset whose productive lifetime or maintenance capacity is physically limited.

Reversibility changes the decision. Deferring a reversible investment may preserve the option to learn about costs and demand. Delaying irreversible prevention may remove a valuable option because physical damage cannot subsequently be undone at acceptable cost. Expected NPV on a static pair of paths does not capture the flexibility of a policy that can respond to new information. A dynamic comparison requires contingent actions, state transitions, and the costs of waiting.

In practical infrastructure evaluation the terminal value may be positive for a useful asset, negative for reclamation or decommissioning, or an uncertain distribution. Capitalizing a terminal value and including the same future service stream separately double counts it. Setting the terminal value to zero at an arbitrary appraisal cutoff can instead hide a large effect. The historical Norwegian long-horizon appraisal review gives an institutional context for treating future effects explicitly, alongside its recommendation of a maturity-dependent rate.

The distinction between appraisal, affordability and delivery is important. A positive social NPV does not supply the funds, institutional capability, land rights, or maintenance commitments necessary to realize an asset. A financing arrangement may be unsustainable even if resource benefits are large. A credible long-horizon decision therefore connects economic valuation to the sequence of obligations that actually produces the predicted outcomes.

The United Kingdom's revised appraisal convention

The United Kingdom supplies an especially useful institutional comparison because its guidance changed during the research year. The 2026 Green Book specifies real rates of 3.00% for years 1–30 and 2.50% for years 31–75, followed by 2.25%. The Treasury review response specifies the final band as years 76–125, explicitly describing the term structure as forward rates. These reported percentages are dated institutional parameters. They are not estimates calculated in the dossier.

The independent review proposed differential treatment of standard and social insurance projects. The government's response accepted the headline 3.00% and declining forward structure while rejecting the proposed lower general rate for social insurance projects. The response also retained the social time preference interpretation while accepting information from the opportunity cost of capital as relevant evidence. This combination illustrates the difference between defining a rate and calibrating its value.

A historically frequently quoted British schedule began at 3.5%; the Norwegian 2012 review documents that earlier British discounting schedule. It should not be substituted for the October 2026 UK guidance in a contemporary appraisal. For a study based on an earlier policy decision, the historical convention can remain relevant. A transparent report therefore states both the decision date and guidance vintage.

The 2026 schedule can be reproduced within its stated bands using D_t=1.03^{-min(t,30)}·1.025^{-max(0,min(t−30,45))}·1.0225^{-max(0,t−75)} for 0≤t≤125. Applying the schedule beyond its documented range requires a separate methodological decision. An agency should also retain sector-specific approved governance where applicable and disclose the reason for a departure from the general convention.

In comparative public economics, the revision reveals that a social discount schedule is simultaneously a numerical representation and a coordinating institution. Changing it recalculates relative project values; it also changes the comparability of proposals across time. The scientific task is to identify the valuation mechanism affected by the revision, distinguish it from revisions to other inputs and preserve the audit trail. It is not to infer social welfare merely from an administrative rate.

United States regulatory analysis and Norwegian public appraisal

In the United States, version control changes the answer to which federal regulatory guidance is operative. The Office of Management and Budget issued its February 2025 reinstatement memorandum, M-25-15, revoking the November 2023 Circular A-4 and reinstating the September 2003 edition. This matters for economic documentation: describing the 2023 revision as the continuing authority without qualification would give an outdated institutional account as of this research snapshot.

The reinstated Circular A-4 (2003) calls for regulatory net-benefit calculations at real rates of 3% and 7%. It distinguishes private capital opportunity costs from consumption effects and allows additional sensitivity at a lower positive rate when important intergenerational consequences are involved. These are regulatory analysis conventions. They are not universal project finance rules, and the document does not establish that either rate represents every person's ethical preferences. Agency-specific statutes or guidance can also affect particular analyses.

The Norwegian expert committee's 2012 cost-benefit analysis report recommended a real risk-adjusted appraisal rate of 4% over the first forty years for ordinary public measures, 3% over the next thirty-five years, and 2% subsequently. This is cited as the historical recommendation in that report rather than an independently verified claim about the legal force of every Norwegian contemporary sectoral rule. Its rationale explicitly connects risk, capital opportunity cost and uncertainties at maturities where a market return cannot readily be secured.

Thus the three documentary examples compare different categories. The UK convention is a contemporary central-government social time preference schedule; the US circular is a reinstated regulatory-analysis dual benchmark; the Norwegian report is a historical expert recommendation using risk-adjusted rates. Their numerical differences cannot be interpreted as a survey ranking national concern for future generations. Coverage, risk treatment, financing, and the institutional status of the documents differ.

For the same illustrative stream the three schedules can nevertheless be applied as mathematical transformations, provided the exercise is explicitly labeled a sensitivity comparison rather than an instruction to apply one country's rules to another's project. Such an exercise reveals which horizons cause differences in valuation. The next section performs this comparison using assumptions that can be fully reproduced.

Numerical comparative appraisal

All amounts in this section are invented real consumption units. The table reports the present value of a fixed receipt of 100 units at each selected horizon. The constant effective annual rates are illustrative; the UK and Norway columns apply the dated forward schedules documented in the preceding discussion, without an additional project-level risk adjustment. The 3% and 7% columns also correspond numerically to the US Circular A-4 benchmark rates. The comparison intentionally holds the payoff constant to isolate the arithmetic; it is not a claim that the institutional approaches value the same risk.

Year Constant 1% Constant 3% Constant 7% UK 2026 Norway 2012 recommendation
30 74.192 41.199 13.137 41.199 30.832
50 60.804 22.811 3.395 25.142 15.499
75 47.413 10.895 0.625 13.562 7.402
100 36.971 5.203 0.115 7.775 4.512
125 28.829 2.485 0.021 4.458 2.750

Calculations use D_t=(1+r)^{-t} for constant rates and the cumulative product of forward factors otherwise. For the UK, the source is the current Green Book schedule; for Norway, the source is the historical committee schedule. Values are rounded only for presentation. In year 30, the UK column equals the constant 3% column by construction. After 30 years the UK factor exceeds that at an unchanging 3%, because subsequent rates are lower. The Norwegian column retains the relatively greater cumulative discount from its initial forty years. Its lower final forward rate does not imply that its earlier discounting disappears.

Return to the 100-unit cost and 400-unit benefit after fifty years. The NPV values at constant 1%, 3%, and 7% are 143.22, -8.76, -86.42 units respectively. At the UK schedule the NPV is 0.57 units, and at the illustrative application of the Norwegian historical schedule it is -38.01. All five calculations are generated directly from the same undiscounted stream. The rate choice alone reverses the sign for some comparisons.

For an alternative investment producing the same 400 units at year 10, the time pattern changes dramatically. Under any of the positive rates in the table, the earlier receipt has a larger present value. If it has higher current cost, the correct ranking requires discounting both costs and benefits, not simply selecting the option with the shortest delivery time. Where investments compete for an indivisible budget, comparing separate benefit-cost ratios can also fail to maximize total welfare; the feasible portfolio must be considered.

The table quantifies a central conclusion: disagreement about remote values may be orders of magnitude greater than disagreement about near-term ones even when all analysts use internally consistent arithmetic. A report that conveys long-horizon valuations without the exact time pattern and a sensitivity range conveys too little information for independent scrutiny.

Empirical identification and the limits of expert estimates

Economics offers several distinct empirical routes to discount-rate inputs. Market yields reveal prices for traded claims in specific institutional settings. Household behavior reveals choices within constraints. Long-run consumption data inform growth and uncertainty. Expert surveys document the beliefs or recommendations of a specified respondent population. None directly observes the complete intergenerational social welfare function.

Newell and Pizer's (2003) interest-rate uncertainty analysis uses a long US historical series and contrasting stochastic processes to estimate distant-future valuations. A crucial issue is distinguishing persistent shocks from mean reversion. Processes that appear similar over the observed sample can imply substantially different long-horizon predictions. Structural changes in monetary institutions, inflation measurement, wars, capital mobility, and demographic conditions further restrict extrapolation.

The survey in Drupp and colleagues (2018), Discounting Disentangled, reports responses from more than 200 experts and a striking degree of acceptance of a 2% median risk-free rate, despite disagreement on its foundations. That result describes an expert sample under the survey's framing. It is not a probability distribution over the true rate, an estimate of public voting preferences, or a proof that 2% is a universal ethical value. In the earlier gamma discounting study, Weitzman (2001) uses expert rate opinions to demonstrate the implications of rate dispersion. Aggregation of expert judgments still requires a separate decision about whether disagreement is modeled as state uncertainty or competing normative commitments.

An empirical study of the effects of mandated discount-rate changes would ideally construct a panel of comparable proposals before and after a reform, with project timing, sector, financing, and risk characteristics. A change in the rate may change which proposals are submitted, not just which are accepted. Cost escalation, political priorities, and budget rules may change simultaneously. An interrupted time-series or comparison design needs credible counterfactual assumptions; a recorded association between a rate change and a project approval count is not sufficient causal evidence.

The present monograph does not estimate such a model or claim to have accessed raw microdata or replication datasets. It identifies the measurements required to distinguish mechanisms and uses reproducible illustrations to show what follows under stated assumptions. Future empirical claims about policy outcomes must be supported by project-level source material and identification strategies rather than extrapolated from the mathematical examples.

Dynamic consistency and institutional revision

Constant exponential utility weighting has a familiar stationarity property. For a deterministic stream evaluated at t, the weight on a consequence at t+s depends only on s. Conditional rankings of future continuation streams can be preserved over time when preferences, opportunities, information and objectives are stable and the same decision criterion is used. A schedule specified by calendar date, updated information, or persistent uncertain rates can preserve probabilistic coherence through conditional revaluation, even if its initially estimated forward curve declines.

A different problem occurs when every future government reuses a maturity-dependent schedule from zero without conditioning on the original probability model or prior commitments. Let D(t)=1/(1+kt), k>0. The relative weight assigned now to dates T and T+1 is D(T+1)/D(T). If the decision is reconsidered at T, a restarted schedule assigns relative weight D(1)/D(0), generally a different ratio. With unchanged facts, the preferred continuation may reverse. This counterexample concerns reapplication of a non-exponential factor and does not prove that every declining government schedule is dynamically inconsistent.

There are legitimate reasons to revisit an appraisal. New evidence about damages or technology can change expected benefits; improved estimates can change the rate or risk premium; changes in financing can alter displaced capital. These are substantive changes in the decision inputs. Policy reversal generated solely by the movement of the valuation date is analytically different. An institution should record the initial decision criterion and the revision rule to distinguish these mechanisms.

The Gollier–Weitzman uncertainty analysis illustrates why changing valuation perspectives without consistent assumptions can lead to conflicting conclusions. The practical issue is how an authority commits to long-term obligations while retaining the capacity to learn. An appraisal method is incomplete if it gives a present value for every future date but no account of how subsequent decisions will be updated.

In infrastructure and public insurance, commitment can support maintenance obligations that protect future cohorts. A rigid commitment can also preserve an outdated technology. Contracting, periodically revised plans, and contingency budgets are economic mechanisms for managing the tension. They should be evaluated against feasible alternatives and the information available at each decision point.

Distributional accountability and collective decision procedures

A benefit-cost calculation aggregates consequences under an accounting and weighting procedure. It can reveal efficient exchanges, dominated alternatives and sensitivity to parameters. Its role in democratic or constitutional decision making depends on the legitimacy of the representation of those affected. The intergenerational dimension is distinctive because future persons cannot participate in current elections or bargaining as present agents.

Public institutions can use durable fiscal commitments, minimum service standards, reserve policies, long-term assets, and independent evaluation to represent future consequences. Each instrument has a cost and a potential failure mode. An earmarked reserve may preserve resources but reduce flexibility; a legal standard may protect a minimum outcome but create compliance costs; an independent analytic body may improve comparability but cannot settle all ethical disagreement. The economic question is how each instrument changes the feasible paths, information, incentives, and allocations.

A transparent decision dossier should present undiscounted annual streams, real-price conventions, discount factors, incidence by groups and cohorts, nonmonetized effects, funding sources, and residual risks. Analysts should demonstrate whether alternative rate assumptions change the preferred action and whether a distribution-sensitive criterion alters its interpretation. The eventual decision can then be related to the premises that determine the ranking.

The institutional literature supports this approach in distinct ways. The OMB (2003) intergenerational discounting discussion recognizes special issues when future persons are affected, while the Norwegian appraisal committee report situates rate design among broader analytical considerations. The documents supply examples of public procedures, not a single definitive constitution of intergenerational justice.

A further test is the distribution of downside risk. A policy can have the largest expected surplus while exposing a minority or future cohort to severe losses. If rights or minimum standards constrain the permissible options, those constraints should be expressed before maximization. If compensation is relied upon, its actual feasibility and enforcement should be shown. An unexecuted potential transfer has different implications from a guaranteed transfer backed by real resources.

Analytical synthesis and decision robustness

The preceding results support a unified interpretation of the economic problem. Social discounting is an intertemporal component of valuation, conditional on an object, unit, welfare criterion, and feasible path. Its usefulness derives from the conversion of dated consequences into a comparable metric. Its limitations arise when the conversion hides relevant differences between those consequences.

The first implication is the separation of judgments. A welfare judgment asks which feasible distribution of well-being is preferred under declared ethical premises. A finance or opportunity-cost judgment asks which resources are displaced and what alternatives they could have produced. An administrative judgment asks which legally and institutionally prescribed appraisal conventions apply. The distinction made by Goulder and Williams (2012) in their two-rate analysis explains why these judgments may call for different numerical inputs. The study's independent synthesis extends the practical distinction to documentation of the administrative convention.

The second implication is that aggregate rate sensitivity must be accompanied by incidence sensitivity. A lower rate increases the absolute present value of a positive distant effect and of a negative distant effect. If costs are shifted to vulnerable future groups, reducing the rate may reveal burdens that a higher rate obscured. It does not intrinsically create fairness. Conversely, a higher consumption discount rate derived from future prosperity can be compatible with equal utility weights. The social conclusion depends on what happens to whom, in which state, at which date.

The third implication concerns uncertainty. The declining-rate derivative demonstrated earlier follows from persistence and a tilted distribution. Expected growth under uncertainty can similarly lower riskless consumption discounting under certain conditions. A decision dependent on positive covariance with future prosperity can behave differently. These results make project-specific state valuation scientifically preferable where credible information exists. When probability assignments are weak, present separate plausible scenarios rather than assign a false precision to a single tail estimate.

A substantive practical test compares four descriptions of the same proposal: its physical consequences, its real consumption-equivalent streams, its distribution-weighted welfare implications, and its resource-financing account. Differences in rankings are diagnostic. A ranking reversal caused by a real relative price assumption points to a valuation problem; a reversal caused by curvature points to distributional or risk assumptions; a reversal caused by capital displacement points to financing; a reversal caused by an administrative schedule points to the institutional convention. Several mechanisms may interact.

This is an analytical synthesis of established economic distinctions, not a proposed proprietary theory. It does not claim that the four descriptions exhaust all possible ethical criteria. Its scientific contribution lies in making the causal and valuation boundaries explicit, identifying double-counting risks, and supplying reproducible counterexamples to unrestricted claims about the sign or ethical meaning of discounting.

Scientific limitations and open questions

The formal benchmarks in this dossier assume marginal changes, positive consumption, stated functional forms, and specified probabilistic structures. Their conclusions cannot be applied to nonmarginal transformations without checking feasibility and the altered marginal values. Intergenerational welfare principles require normative justification that data alone cannot supply. Present-value results are conditional on which outcomes and persons enter the account.

The documentary comparison is intentionally selective. It includes the current October 2026 UK documents, the February 2025 US reinstatement and 2003 Circular A-4, and the 2012 Norwegian expert recommendation. It does not purport to inventory every national or multilateral development-bank practice. The dated institutional sources establish the described rules and recommendations, not universal applicability or realized policy effects. The expert surveys are interpreted within their stated sample and purpose. Several journal sources are accessible here through verified publisher abstracts or bibliographic records rather than complete subscription text. Claims based on those sources are limited to the available evidence and historical context; all formal derivations and numerical examples are independently reconstructed.

Important unresolved scientific questions concern the empirical persistence of consumption shocks across long horizons, the relationship between individual risk preferences and collective inequality aversion, the specification of relative prices under ecological scarcity, robust treatment of extreme outcomes, and the design of institutional commitments that can incorporate new evidence without arbitrary rediscounting. Population ethics and the valuation of irreversible loss remain contested areas where transparent assumptions are essential.

A full empirical extension would require access to raw project appraisal records, historical series with documented real-rate construction, forecasts with vintage and error distributions, and consistent cohort or regional incidence accounts. A replication of the cited expert surveys would require their underlying questionnaires and data. No such replication or primary data collection is asserted here. These research frontiers do not invalidate the conditional algebra; they define the evidence needed for stronger application.

The dossier's conclusions therefore distinguish exact mathematical implications, document-specific institutional facts, and conditional economic interpretations. For this reason, precision about scope strengthens rather than diminishes the usefulness of a long-horizon study.

Conclusion

Intergenerational economic choice evaluates paths whose consequences are distributed across time, persons and states of the world. Social discounting is one component of this evaluation. Under deterministic additive welfare and constant relative curvature, the marginal consumption rate combines pure utility discounting and growth multiplied by curvature. Under uncertainty, expected marginal utility and covariance determine the value of a contingent payoff. Uncertain persistent rates can generate a declining forward schedule, while independent shocks need not do so.

The historical and institutional literature demonstrates why market valuation, social welfare valuation and public appraisal cannot be identified by a single observed percentage. The United Kingdom's 2026 reform, the United States' reinstated regulatory guidance, and Norway's historical recommendation provide contrasting documented approaches. Their applicability depends on the defined administrative context and risk treatment. Numerical comparisons demonstrate that small differences in effective annual rates can reverse the monetary ranking of remote benefits, and that a forward schedule must be accumulated rather than applied retroactively.

The central economic conclusion is that the quality of an intergenerational decision depends on the coherence of the entire valuation account. Discount factors require an explicit numeraire, benefits and costs require credible counterfactuals, distributional weights require an ethical interpretation, and capital costs require a financing account. Policy-dependent catastrophic risks, changing relative prices and nonmarket consequences require explicit treatment. The analysis supports transparent sensitivity and distributional comparisons over reliance on an isolated headline rate.

The research thereby establishes a rigorous basis for evaluating long-horizon economic arguments while preserving the distinction between mathematical implications, empirical evidence and legitimate collective choice. Its analytical contribution consists in the explicit reconciliation of these established domains and the reproducible comparisons supplied above. No new named economic theory or claim of priority is advanced.

References

  • Ramsey, F. P. (1928). A Mathematical Theory of Saving. The Economic Journal, 38(152), 543–559. doi:10.2307/2224098. Source
  • Frederick, S., Loewenstein, G., & O’Donoghue, T. (2002). Time Discounting and Time Preference: A Critical Review. Journal of Economic Literature, 40(2), 351–401. doi:10.1257/002205102320161311. Source
  • Arrow, K. J., Cropper, M. L., Gollier, C., Groom, B., Heal, G. M., Newell, R. G., Nordhaus, W. D., Pindyck, R. S., Pizer, W. A., Portney, P. R., Sterner, T., Tol, R. S. J., & Weitzman, M. L. (2014). Should Governments Use a Declining Discount Rate in Project Analysis? Review of Environmental Economics and Policy, 8(2), 145–163. doi:10.1093/reep/reu008. Source
  • Gollier, C., & Weitzman, M. L. (2010). How Should the Distant Future Be Discounted When Discount Rates Are Uncertain? Economics Letters, 107(3), 350–353. doi:10.1016/j.econlet.2010.03.001. Source
  • Weitzman, M. L. (2001). Gamma Discounting. American Economic Review, 91(1), 260–271. doi:10.1257/aer.91.1.260. Source
  • Newell, R. G., & Pizer, W. A. (2003). Discounting the Distant Future: How Much Do Uncertain Rates Increase Valuations? Journal of Environmental Economics and Management, 46(1), 52–71. doi:10.1016/S0095-0696(02)00031-1. Source
  • Drupp, M. A., Freeman, M. C., Groom, B., & Nesje, F. (2018). Discounting Disentangled. American Economic Journal: Economic Policy, 10(4), 109–134. doi:10.1257/pol.20160240. Source
  • Goulder, L. H., & Williams, R. C., III. (2012). The Choice of Discount Rate for Climate Change Policy Evaluation. Resources for the Future, Discussion Paper 12-43 (also NBER Working Paper 18301). Source
  • Stern, N. (2007). Economics, Ethics and Climate Change. In The Economics of Climate Change: The Stern Review, 25–45. Cambridge University Press. doi:10.1017/CBO9780511817434.006. Source
  • Nordhaus, W. D. (2007). A Review of the Stern Review on the Economics of Climate Change. Journal of Economic Literature, 45(3), 686–702. doi:10.1257/jel.45.3.686. Source
  • Chichilnisky, G. (1996). An Axiomatic Approach to Sustainable Development. Social Choice and Welfare, 13(2), 231–257. doi:10.1007/BF00183353. Source
  • HM Treasury. (2026). The Green Book: Central Government Guidance on Appraisal and Evaluation. October 2026 edition. London: HM Treasury. Discounting section and Table 8. Source
  • HM Treasury. (2026). Green Book Discount Rate Review: HM Treasury Response. October 2026. London: HM Treasury. Responses R.7, R.9, R.17–R.19. Source
  • Norwegian Ministry of Finance, expert committee. (2012). Cost-Benefit Analysis. Official Norwegian Reports NOU 2012:16, chapter 5, The Social Discount Rate. Oslo: Government of Norway. Source
  • US Office of Management and Budget. (2003). Circular A-4: Regulatory Analysis. September 17, 2003. Washington, DC: Executive Office of the President, sections E and F, pp. 33–36. Source
  • US Office of Management and Budget. (2025). Memorandum M-25-15: Recission and Reinstatement of Circular A-4. February 12, 2025. Washington, DC: Executive Office of the President. Source
  • Arrow, K. J., Cropper, M. L., Gollier, C., Groom, B., Heal, G. M., Newell, R. G., Nordhaus, W. D., Pindyck, R. S., Pizer, W. A., Portney, P. R., Sterner, T., Tol, R. S. J., & Weitzman, M. L. (2012). How Should Benefits and Costs Be Discounted in an Intergenerational Context? The Views of an Expert Panel. Resources for the Future, Discussion Paper 12-53, December 2012. Source
  • Ramsey historical context: Attanasio, O. P. (2015). Frank Ramsey's A Mathematical Theory of Saving. The Economic Journal, 125(583), 269–294. doi:10.1111/ecoj.12229. Source