The Interest Rate of Morality: The Social Discount Rate and What the Future Is Worth to Us

The Hook: A Number Bigger Than Any Debate About It

Imagine a decision you make this year. It costs money today – say, rebuilding a data centre, reinforcing a dyke, forgoing a cheap energy source. The benefit arrives in a hundred years: one million euros of damage that then does not occur.

What is that benefit worth today?

The question sounds technical. It is the arithmetic that sits at the end of every cost-benefit analysis, and it has a simple answer: divide the future amount by the compounding factor over the period. Except that everything depends on which rate you put in. At 1.5 percent per year, that avoided million euros is worth roughly 225,600 euros today. At 5 percent, it is 7,600 euros. The same damage, the same physics, the same figure a hundred years out – and a factor of almost thirty between the two answers.

Over two hundred years, the factor of thirty becomes a factor of almost nine hundred: 50,900 euros against 58 euros. At 5 percent over three hundred years, the present value of a million falls below one euro.

This is not a quibble of economics. It is the point at which an ethical decision – how much do people who come after us count? – migrates into models, agency guidance and regulatory impact assessments in the shape of a parameter, and becomes invisible there. The number is called the social discount rate. It is, as I would put it, the most powerful ethical lever in modern policy advice – and the one most often treated in public debate as a purely technical question, even though specialists argue fiercely about it.

This article follows that parameter from its birth in a 1928 paper through the most famous controversy in climate economics to the numbers that sit in American and German agency documents today – and to the day a constitutional court built a fundamental right out of a structurally related consideration.

Part 1: The Core Concept – How to Cast an Ethics as a Percentage

Two Reasons to Count the Future for Less

Discounting conflates at least two entirely different ideas, and a large part of the confusion comes from not separating them. (That it does not stop at two in practice is shown in Part 4.)

The first idea is descriptive: future people will probably be richer. For a richer person, an additional euro is worth less than it is for a poor one – the diminishing marginal utility of consumption. So whoever values a damage a hundred years out is valuing it for people to whom that same euro means less. This is not disdain for the future; it is a statement about wealth.

The second idea is normative: you weight well-being less simply because it arrives later. This is the pure rate of time preference – usually δ in the literature, PRTP in English. Here there is no excuse about wealth. Setting δ greater than zero says: suffering in the year 2150 counts for less than the same suffering today, because it happens in 2150.

Frank Plumpton Ramsey put this, in 1928, into the model that remains the starting point of the debate: "A Mathematical Theory of Saving" (The Economic Journal 38, no. 152, 543–559). In the main part of his model he deliberately set the pure rate of time preference to zero. His justification has become one of the best-known remarks in economics: he calls discounting later enjoyments "a practice which is ethically indefensible and arises merely from the weakness of the imagination".

In the second section of the paper, Ramsey does also work through a positive constant rate, with the illustrative value "about 2 per cent" – without defending it ethically. What stands is this: Ramsey's baseline was δ = 0, and that was an ethical decision, not an empirical one.

The Ramsey Equation

Both ideas together give the equation that has framed the debate for nearly a century:

r = δ + η · g

• r is the social discount rate.
• δ is the pure rate of time preference – the normative part.
• g is the expected growth rate of consumption per head.
• η is the elasticity of marginal utility of consumption – a measure of how quickly the value of an additional euro falls as wealth rises.

In the standard model, η is simultaneously the measure of inequality aversion and of risk aversion. Empirically, however, these three roles come apart. Atkinson, Dietz, Helgeson, Hepburn and Sælen note in "Siblings, not triplets" (2009) that while the three concepts are necessarily identical in the standard model, this need not be so – their survey evidence suggests that people do not treat them as the same parameter, and that preferences over risk, inequality and time are only weakly correlated.

The equation is elegant because it exposes the ethics instead of hiding it: of its three parameters only one is a forecast; one is a pure value judgement, and one – η – sits between the two. And it is dangerous because it is so handy that you can type it into any spreadsheet without noticing that you have just voted on the rights of unborn people.

A numerical example makes the leverage visible. With δ = 0.1 percent, η = 1 and g = 1.5 percent, the equation gives 1.6 percent. With δ = 1.5 percent, η = 2 and g = 2 percent, it gives 5.5 percent. Both parameter sets lie within the range of values defended in the literature. The difference in the result – 1.6 against 5.5 percent – decides trillions.

Why the Leverage Is So Brutal

Compounding is exponential, and exponential functions forgive no small deviation over long periods. The table below shows the present value of one million euros of damage occurring in n years. All values are simple discounting, 1,000,000 / (1 + r)^n, recomputed by me:

Read the right-hand column again. At 5 percent, the one-off damage of a species extinction, the loss of a coastal city or a drought three hundred years from now has shrunk, in the calculation, to under one euro – practically indistinguishable from zero. (A permanent loss would be a stream of payments and would fall more slowly, but dramatically all the same.) The same model, the same formula, run at 0.1 percent, treats the damage almost as if it happened tomorrow.

From this follows an uncomfortable insight – but one not to be overstated. Whoever calculates over long horizons has pulled the single largest lever when choosing the discount rate, but not the only one: in the recalculation by Rennert and co-authors, which Part 5 returns to, updating the damage functions alone raised the figure by 81 percent before the rate was changed at all.

Part 2: Stern versus Nordhaus – The Clash That Was Really About Ethics

The Stern Review, 2006

On 30 October 2006, Nicholas Stern delivered a report on the economics of climate change to the British Treasury. His finding was radical: immediate, decisive action is far cheaper than inaction.

The parameters with which the Review computed this are well documented. The House of Commons Library Standard Note SN/EP/4739 ("The Stern Review and the discount rate") quotes Stern's own FAQ: "we assume a low rate of pure time preference, 0.1%, which corresponds with a 90% probability of humanity surviving a 100-year period" – so δ = 0.1 percent, justified not by impatience but by the risk that humanity will not be there in a hundred years. And on the second quantity: "We use an elasticity of one, in line with some empirical estimates" – η = 1.

The rest is arithmetic. The note reproduces a contemporary summary from the Financial Times of 15 November 2006: "If we assume a rate of growth of consumption of 1.5 per cent a year, the discount rate would then be just 1.6 per cent a year. With 2 per cent consumption growth, it would be 2.1 per cent." (The note names no author for that article.)

One qualification matters here: the Review itself computed stochastically with the PAGE2002 model and did not use one rate that could be quoted. The 1.6 percent is an illustration, not a model constant.

With these parameters the Review arrived at damages on the business-as-usual path which, in Stern's formulation as quoted in the note, were "equivalent to at least a 5% loss in consumption", "based on a narrow definition of risks and impacts, and up to 20% if a broader range of risks and impacts are considered".

Stern was no outsider at that point: knighted as early as 2004, he headed the British Government Economic Service from 2003 to 2007; in 2007 he entered the House of Lords as Baron Stern of Brentford.

Nordhaus's Objection

William Nordhaus – who had been developing the DICE model since the early 1990s – answered quickly: first in December 2006 as NBER Working Paper 12741, then published in September 2007 in the Journal of Economic Literature as "A Review of the Stern Review on the Economics of Climate Change" (vol. 45, no. 3, 686–702, DOI 10.1257/jel.45.3.686). The core of his objection is in the abstract, and it is remarkably terse:

"An examination of the Review's radical revision of the economics of climate change finds, however, that it depends decisively on the assumption of a near-zero time discount rate combined with a specific utility function. The Review's unambiguous conclusions about the need for extreme immediate action will not survive the substitution of assumptions that are consistent with today's marketplace real interest rates and savings rates."

Three things matter about this – and the debate is frequently summarised wrongly on all three.

First: Nordhaus did not dispute the climate physics. He did attack the damage estimates substantially – they were "much higher than the standard estimates in the impact literature", and the Review "drew selectively from studies, emphasizing those with high damage estimates" – but he named discounting as the central driver of the result.

Second: his main argument was one of consistency and calibration, not an ethical one. If societies really acted with a 0.1 percent pure rate of time preference, they would have to save and invest far more than they do; observed real interest rates and savings rates say otherwise. He does name ethical counter-positions, though: he calls consumption sacrifice by today's poorer generations for the benefit of richer future ones "hardly … ethically compelling", and he lists rival frameworks (capital preservation, Rawls, minimax) with the conclusion that alternative ethical perspectives are possible.

Third – and this is the most commonly missed point: Nordhaus did not dispute δ alone, but the combination of δ and η. In his own recalibration he keeps Stern's 0.1 percent pure rate of time preference and instead sets η to 2.25 – and thereby obtains a policy path close to the standard model. The controversy was therefore not about the parameter usually attributed to him.

What the Clash Really Was

It was not about whether climate change is real or dangerous. Both sides affirmed that. It was about a question as old as moral philosophy: should a policy model represent the values people actually display (descriptive), or the ones they ought to hold (prescriptive)?

Stern's position: a government deciding for centuries must not adopt the impatience of consumers at a savings account. Nordhaus's position: a model whose parameters are compatible with no observable behaviour does not yield testable recommendations but dresses a prior decision up as a calculation.

The core of the dispute – the choice of δ – cannot be settled by better data; it is a dispute about the kind of question being asked. The side fronts can be: η is partly estimable empirically, and consistency with real interest rates and savings rates is a matter of measurement.

I am of the opinion that both are right within their own frame – and that this is precisely what is disquieting. In 2018 Nordhaus was awarded the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel, in equal shares with Paul Romer, in his case "for integrating climate change into long-run macroeconomic analysis". Neither refuted the other.

Part 3: When Disagreement Itself Becomes the Argument

There is a third way, and it is the most elegant idea in this entire field.

In 2001 Martin Weitzman published "Gamma Discounting" in the American Economic Review (vol. 91, no. 1, 260–271, DOI 10.1257/aer.91.1.260). He ran a survey of economists and drew a conclusion from it that sounds paradoxical at first. From the abstract:

"A numerical example is constructed from the results of a survey based on the opinions of 2,160 economists. The main finding is that even if every individual believes in a constant discount rate, the wide spread of opinion on what it should be makes the effective social discount rate decline significantly over time."

The mechanism is, once seen, plain – what is contested is only whether one should average present values or terminal values. Weitzman's proposal was to average the individual discount factors, not the rates. Suppose half the profession holds 1 percent to be right and the other half 6 percent. Over short horizons the two camps are not far apart; after one year the effective rate is about 3.4 percent. Over long horizons the 6 percent camp drives its present value practically to zero – after three hundred years the averaged present value comes almost entirely from the 1 percent camp, and the effective rate has fallen to about 1.2 percent.

What matters is the direction of that fall and its limit. The averaged present value is by construction the mean of the individual present values and therefore always lies between the lowest and the highest; it can never exceed that of the most patient participant. Gollier and Weitzman put it in 2010 as the long-run discount rate declining over time toward its lowest possible value – toward it, not below it.

In other words: uncertainty about the right discount rate is no reason for inaction – if you average present values, it becomes an argument for giving the distant future more weight than a constant rate at the average of the quoted rates would give it.

That this averaging is the right one was, however, contested for more than a decade. Weitzman had not initially set out his rationale in full; Christian Gollier showed that averaging the expected terminal value leads to the opposite result – rising rather than falling rates. This tension became known as the Weitzman-Gollier puzzle. Gollier and Weitzman jointly declared it resolved in 2010, in favour of declining rates, but under one condition: that agents optimise their consumption plans. And Freeman and Groom still consider the practical use contentious: if the spread of opinion reflects ethical disagreement, a declining path is defensible; if it reflects merely forecasting error about the same rate, the path is much flatter, and standard approaches potentially exaggerate the decline.

In 2013 an unusually distinguished group of authors took this subject to American policymakers in Science: Kenneth J. Arrow, Maureen L. Cropper, Christian Gollier, Ben Groom, Geoffrey M. Heal, Richard G. Newell, William D. Nordhaus, Robert S. Pindyck, William A. Pizer, Paul R. Portney, Thomas Sterner, Richard S. J. Tol and Martin L. Weitzman, "Determining Benefits and Costs for Future Generations" (Science 341, no. 6144, 349–350, 26 July 2013, DOI 10.1126/science.1235665). The summary held in Crossref reads: "The United States and others should consider adopting a different approach to estimating costs and benefits in light of uncertainty." Note: I could not open the full text of the piece in this run; the frequently cited concrete recommendation of a declining discount rate schedule is therefore not carried here as a finding of my own (see Fact Check). What is documented is this: Nordhaus and Weitzman, who six years earlier stood on opposite sides of the Stern controversy, signed the same appeal to calculate differently under uncertainty.

Part 4: What the Profession Actually Recommends

An obvious objection runs: if economists disagree so fundamentally, the instrument is no good. That objection is – empirically – weaker than one might think.

Moritz A. Drupp, Mark C. Freeman, Ben Groom and Frikk Nesje measured this in 2018 in the American Economic Journal: Economic Policy ("Discounting Disentangled", vol. 10, no. 4, 109–134, DOI 10.1257/pol.20160240). From the abstract:

"We surveyed over 200 experts to disentangle disagreement on the risk-free SDR into its component parts, including pure time preference, the wealth effect and return to capital. We show that the majority of experts do not follow the simple Ramsey Rule, a widely-used theoretical discounting framework, when recommending SDRs. Despite disagreement on discounting procedures and point values, we obtain a surprising degree of consensus among experts, with more than three-quarters finding the median risk-free SDR of 2 percent acceptable."

Three findings worth remembering.

First: the median for the risk-free rate is 2 percent. That is considerably closer to Stern than to the rates agencies used for a long time – around 7 percent at the US Office of Management and Budget, 3.5 percent in the British Green Book.

Second: more than three-quarters of respondents consider this value acceptable – including many who would name a different figure themselves. That is, as I read it, a stronger measure of consensus than agreement on a point value, and the more relevant one for policy. Drupp and co-authors themselves speak of "a surprising degree of consensus".

Third – and this is usually overlooked: the majority of experts do not follow the simple Ramsey Rule. The equation from Part 1 is the shared vocabulary of the field, but not the procedure by which the specialists form their own recommendations. The survey decomposes the justifications into pure time preference, the wealth effect and the return to capital – so in practice it does not stop at the two ideas from Part 1.

Part 5: The Number in Official Hands

With this, the debate leaves the seminar room. The social discount rate today sits inside a single indicator that appears in regulatory impact assessments on both sides of the Atlantic: the climate cost per tonne of CO2, or in American terminology the social cost of carbon.

The United States

Kevin Rennert and co-authors published a recalculation in Nature in 2022 ("Comprehensive evidence implies a higher social cost of CO2", vol. 610, no. 7933, 687–692). From the abstract:

"Our preferred mean SC-CO2 estimate is $185 per tonne of CO2 ($44–$413 per tCO2: 5%–95% range, 2020 US dollars) at a near-term risk-free discount rate of 2%, a value 3.6 times higher than the US government's current value of $51 per tCO2."

The comparison value of 51 dollars rested on a constant rate of 3 percent. Where does the jump come from? The paper decomposes it – though not starting from 51 but from the DICE-2016R baseline of 44 dollars: an updated climate module, socioeconomics and discounting approach raise it to 59 dollars, the new sectoral damage functions to 80 dollars, and the move to a 2 percent near-term rate finally to 185 dollars. The discount rate is thus, in the authors' words, "the single largest contributor"; the updated damage functions are the second largest. Pointing to the rate alone is therefore a simplification – but no other single assumption comes close to it.

The US Environmental Protection Agency followed suit. Its report "Report on the Social Cost of Greenhouse Gases: Estimates Incorporating Recent Scientific Advances" (November 2023, Docket EPA-HQ-OAR-2021-0317) names, in Table ES.1 – which also covers methane and nitrous oxide – three explicit near-term target rates, from which the rate declines over the time horizon. The agency thereby implements Weitzman's idea from Part 3 operationally: uncertainty about the starting rate is represented through three target rates of 1.5, 2.0 and 2.5 percent, calibrated to the decline in the certainty-equivalent discount rate, yielding dynamic discount rate paths. The report states explicitly that it uses "the Ramsey (1928) discounting formula". The CO2 values in 2020 dollars, rounded to two significant figures:

For the same emission year 2020 the range runs from 120 to 340 dollars – a factor of 2.8, from the choice of near-term rate alone. Note what this means: the agency publishes not one number but three – each with its own uncertainty distribution (the report shows the full distributions and notes that they have long right tails) and with the parameter disclosed. That is methodologically honest and politically awkward, because each side can pick its column.

A word of caution: the EPA's 190 dollars at 2.0 percent and Rennert and co-authors' 185 dollars resemble each other, but they are not the same thing – and their closeness is not independent corroboration. The EPA averages over three damage modules, and one of them is exactly the GIVE model of Rennert and co-authors, which the report cites explicitly as "developed under RFF's Social Cost of Carbon Initiative". The closeness of the numbers is therefore partly shared provenance under different aggregation.

Germany

The German Environment Agency (Umweltbundesamt, UBA) takes a different route, and one remarkably transparent from an ethical point of view. It names no single recommended rate at all, but two – and states what distinguishes them.

Its "Methodenkonvention 3.1 zur Ermittlung von Umweltkosten – Kostensätze" (status 12/2020, Dessau-Roßlau, authors Astrid Matthey and Björn Bünger) reads:

"Wir empfehlen die Verwendung eines Kostensatzes von 195 €2020 / t CO2 äq für das Jahr 2020 bei einer Höhergewichtung der Wohlfahrt heutiger gegenüber zukünftigen Generationen und eines Kostensatzes von 680 €2020 / t CO2 äq bei einer Gleichgewichtung der Wohlfahrt heutiger und zukünftiger Generationen."

(We recommend using a cost rate of €195 (2020 prices) per tonne of CO2 equivalent for the year 2020 where the welfare of present generations is weighted more heavily than that of future generations, and a cost rate of €680 (2020 prices) per tonne where the welfare of present and future generations is weighted equally.) And immediately after: "Zusätzlich empfehlen wir eine Sensitivitätsanalyse mit dem jeweils anderen Wert." (We additionally recommend a sensitivity analysis using the other value.)

The accompanying table assigns the values to the parameters: 195 euros corresponds to a pure rate of time preference of 1 percent, 680 euros to one of 0 percent. For 2050 the values are 250 euros and 765 euros respectively.

The factor between the two recommendations is 3.5. In that table it arises exclusively from a single percentage point of pure time preference – all other specifications are stated jointly for both rows. In other words, from precisely the quantity no dataset can settle.

The English-language successor version now available ("Methodological Convention 3.2 for the Assessment of Environmental Costs – Value Factors", version 10/2024) gives €300 (2024 prices) per tonne of CO2 equivalent at a 1 percent pure rate of time preference and €880 (2024 prices) at 0 percent, and for 2050 €435 and €1,080 respectively. The values come from a new model (GIVE instead of FUND) and from several further adjustments – equity weighting, a lower value of a statistical life, price and purchasing-power adjustment – and are therefore not a simple continuation of the 2020 figures. What is decisive is the justification the convention itself articulates:

"A positive PRTP can thus be interpreted as prioritizing the needs of current generations over those of future generations, and consequently presents a value judgement."

That is, in an official document, Ramsey's 1928 question in plain words: whoever sets δ > 0 makes a value judgement. Not a measurement.

Part 6: When the Law Weighed In

On 24 March 2021 the First Senate of the German Federal Constitutional Court decided, in cases 1 BvR 2656/18, 1 BvR 78/20, 1 BvR 96/20 and 1 BvR 288/20, something that can be read in the language of fundamental rights as an answer to the same question – without the court mentioning discounting anywhere. Headnote 4 of the order states this in three sentences; the second reads:

"Subjektivrechtlich schützen die Grundrechte als intertemporale Freiheitssicherung vor einer einseitigen Verlagerung der durch Art. 20a GG aufgegebenen Treibhausgasminderungslast in die Zukunft."

The court's English press release (no. 31/2021 of 29 April 2021) puts the same point this way: "As intertemporal guarantees of freedom, fundamental rights afford the complainants protection against comprehensive threats to freedom caused by the greenhouse gas reduction burdens that are mandatory under Art. 20a GG being unilaterally offloaded onto the future." And on the doctrinal construction: "The challenged provisions have an advance interference-like effect (eingriffsähnliche Vorwirkung) on the freedom comprehensively protected by the Basic Law."

The operative part was concrete. Under point 2 of the order, "§ 3 Absatz 1 Satz 2 und § 4 Absatz 1 Satz 3 Bundes-Klimaschutzgesetz vom 12. Dezember 2019 (Bundesgesetzblatt I Seite 2513) in Verbindung mit Anlage 2 […] mit den Grundrechten unvereinbar, soweit eine den verfassungsrechtlichen Anforderungen nach Maßgabe der Gründe genügende Regelung über die Fortschreibung der Minderungsziele für Zeiträume ab dem Jahr 2031 fehlt" – the named provisions of the Federal Climate Change Act are incompatible with fundamental rights in so far as they lack a constitutionally adequate rule for updating the reduction targets for periods from 2031 onwards. Under point 4, the legislature was obliged to remedy this "spätestens bis zum 31. Dezember 2022" – by 31 December 2022 at the latest.

One should not claim too much here. The court set no discount rate and prescribed no economic method. What it did was to attribute to the unilateral offloading of burdens onto the future an "advance interference-like effect" on future freedom – a deliberately restrained doctrinal construction, not a completed interference – and thereby to make it something requiring justification. A high discount rate in a spreadsheet does something structurally similar: it offloads no physical reduction budget but weights damages – only without any burden of justification at all.

Frameworks to Take Away

The three questions behind every discount rate

The map of positions

The Central Takeaway

Four points that carry beyond climate economics.

1. Find the parameter that carries the ethics. In models with a long horizon there is almost always a place where a value judgement is entered as a number. In climate models it is δ. In an investment appraisal it is the hurdle rate; in a risk assessment it is the probability one calls "conservative"; in an architecture decision it is the period over which operating costs are set against build costs. Whoever finds that place understands the result before working through the arithmetic.

2. Separate the descriptive from the normative – even when the formula conflates them. The Ramsey equation adds a forecast (g) and a value judgement (δ) into a single number, with η as the hybrid in between. Afterwards it is no longer visible which part came from where. This is a typical error in discussions about discounting: arguing empirically about a quantity that has a decision inside it.

3. Disagreement about the rate is no reason to ignore the distant future – arithmetically it is the opposite. That is Weitzman's point: average present values rather than rates, and the low rates dominate the long horizon. Translated: when the discounting of a long-term risk is disputed, "we don't know, so let's leave it out" is the worst-justified answer – under uncertainty about the rate, the expected value of the distant future is higher, not lower, than an average rate implies. That this mechanism is the right kind of aggregation was itself contested, though (Part 3), and how steeply the rate falls depends on whether the spread reflects ethical disagreement or merely forecasting error.

4. Quote two numbers, not one. The methodologically best thing the agencies in this story have done, in my view, is not to have found the right rate – it is to publish several rates in parallel and to say which difference in the world corresponds to the difference between columns. The next time you put forward a long-horizon number – total cost of ownership, payback period, technical debt, risk provisioning – put the second calculation with the other parameter beside it. Not as an uncertainty band, but as disclosure of the value judgement.

Cross-References in the Vault

• The Billions Barely Alive: Parfit's Repugnant Conclusion and the Ethics of Future Generations – the other great question in the ethics of future generations: not how heavily we weight them, but how many there should be.
• The Child in the Pond: How a Three-Page Essay by Peter Singer Launched a Global Movement — and Why the Question of How Far Our Duties Reach Echoes All the Way into the Boardrooms of Tech Billionaires – spatial rather than temporal distance; on this reading the discount rate is a distance discount in time.
• Behind the Veil of Ignorance: How Rawls Reinvented Justice – and Why His Thought Experiment Still Echoes in the Ethics of Artificial Intelligence – behind the veil you also do not know which generation you wake up in; Nordhaus names Rawls explicitly as an ethical alternative.
• The Marshmallow Promise: How a Candy Test Redefined Willpower for Half a Century – and Was Then Put to the Test Itself – individual time preference, to be kept sharply distinct from the social discount rate.
• Losses Loom Larger: How Prospect Theory Dethroned Rational Man – and Why Its Most Famous Principle Is Now on Trial Itself – how people actually decide about time and risk; Nordhaus's descriptive argument lives on findings like these.
• The Gentle Push: Why a Pre-Ticked Box Achieves More Than Any Argument — and Whether It Is Permissible to Nudge People Toward Their Own Good – a second case in which a technical setting carries a value judgement.
• Who Goes First: Triage, the Constitution, and the Calculation Nobody Wants to Make – allocation under scarcity, there spatial and acute, here temporal and slow.

Something to Think About

Suppose you had to set a single percentage rate for your own organisation, to be used in evaluating every decision with effects beyond thirty years – and suppose you had to publish that rate with one sentence of justification that may not refer to market interest rates.

Which number do you write down? And to whom would you have to justify yourself if it is too high – given that those affected cannot object today?

Fact Check

For this article, 117 checkable claims were extracted (49 hard facts, 19 verbatim quotations, 18 source attributions, 31 synthesis claims) and checked by a separate review pass against the primary and publisher sources opened in this run. 4 claims were false, 33 were weakened, and 3 were not verifiable against a source – 37 corrections in total. Every table value (the discounting arithmetic, the EPA's Table ES.1, the UBA cost rates) and every checked quotation held up character for character; the weakness lay almost entirely in the class of synthesis claims. The most important corrections:

1. The conclusion drawn from Weitzman was mathematically false. The draft claimed that averaging makes the distant future more valuable "than any individual participant in the discussion would". But the averaged present value lies by construction between the lowest and the highest and can never exceed that of the most patient participant; Gollier and Weitzman (2010) frame the limit as convergence toward the lowest possible rate. The claim now correctly reads: higher than a constant rate at the average of the quoted rates. The Weitzman-Gollier puzzle, which the draft omitted, was also added – the choice of aggregation was contested for a decade.
2. Nordhaus did not dispute the parameter δ, but the combination of δ and η. The draft compressed his critique into "he disputed the parameter". In his own recalibration, however, Nordhaus retains Stern's 0.1 percent pure rate of time preference and changes η to 2.25. Also corrected: Nordhaus attacked the Review's damage estimates substantially ("much higher than the standard estimates in the impact literature") and does name ethical counter-positions of his own – the draft had presented his argument as "not an ethical one".
3. The EPA rates are not constant rates but near-term target rates on declining paths. The draft presented the three columns as three fixed rates and thereby missed its own point from Part 3: the EPA implements Weitzman's declining rates operationally. Also corrected: the closeness of the EPA's 190 dollars to Rennert's 185 dollars is not independent corroboration – one of the EPA's three damage modules is Rennert's GIVE model.

Further corrections concern: the attribution of the 1.6 percent calculation (it comes from a Financial Times article of 15 November 2006 reproduced in the Library note, not from the note itself or from Stern's FAQ); the citation point at the Federal Constitutional Court (the quoted sentence is the second of three sentences of headnote 4, not the headnote); the claim that the court declared an "interference requiring justification" (it says expressly "advance interference-like effect" – an internal contradiction with the wording correctly quoted two paragraphs earlier); Stern's biography (knighted in 2004, that is before the Review; life peer only in 2007); Ramsey's role (not "the first", and in section II he does work through a positive rate); the equation of η with inequality aversion (holds only in the standard model and is empirically contested); the decomposition of the jump from 51 to 185 dollars (the discount rate is "the single largest contributor", not the primary cause, and the decomposition starts at 44 dollars); the claim that agencies long used "4 to 5 percent" (in fact around 7 percent at the OMB, 3.5 percent in the British Green Book); and a series of overly strong generalisations in "The Central Takeaway" ("in every model", "the most common error", "the only answer that is demonstrably false", "literally worth nothing").

The 1.6 percent formulation and the 5 percent damage quotation were checked against a mirror copy of the House of Commons note, because the original became inaccessible after the first retrieval. The full text of the Science piece by Arrow and co-authors remained inaccessible; the recommendation of a declining discount rate schedule often attributed to it is therefore not carried as a finding in this article.

Note: automated AI verification against primary sources; no human expert review.

Sources and Further Reading

• F. P. Ramsey, "A Mathematical Theory of Saving", The Economic Journal 38, no. 152, 543–559 (1 December 1928), DOI 10.2307/2224098 – origin of the Ramsey equation; bibliographic data verified via the publisher page: https://academic.oup.com/ej/article/38/152/543/5287992
• "Frank Ramsey's Contributions to Economics", Stanford Encyclopedia of Philosophy – placement of the paper and verification of the quoted passage: https://plato.stanford.edu/entries/ramsey-economics/
• Dominic Webb, "The Stern Review and the discount rate", House of Commons Library Standard Note SN/EP/4739 (status 9 June 2008) – quotes Stern's own FAQ on δ = 0.1 % and η = 1 as well as the Financial Times summary: https://researchbriefings.files.parliament.uk/documents/SN04739/SN04739.pdf
• William D. Nordhaus, "A Review of the Stern Review on the Economics of Climate Change", Journal of Economic Literature 45, no. 3, 686–702 (September 2007), DOI 10.1257/jel.45.3.686: https://www.aeaweb.org/articles?id=10.1257/jel.45.3.686 – earlier version as NBER Working Paper 12741 (December 2006): https://www.nber.org/papers/w12741.pdf
• Martin L. Weitzman, "Gamma Discounting", American Economic Review 91, no. 1, 260–271 (March 2001), DOI 10.1257/aer.91.1.260: https://www.aeaweb.org/articles?id=10.1257/aer.91.1.260
• Christian Gollier and Martin L. Weitzman, "How Should the Distant Future Be Discounted When Discount Rates Are Uncertain?" (2010) – resolution of the Weitzman-Gollier puzzle under optimising consumption plans: https://stafforini.com/works/gollier-2010-how-should-distant/
• Mark C. Freeman and Ben Groom on the practical use of declining rates and whether the spread reflects ethical disagreement or forecasting error: https://eprints.whiterose.ac.uk/id/eprint/111966/1/FreemanGroomManuscript.pdf
• Giles Atkinson, Simon Dietz, Jennifer Helgeson, Cameron Hepburn and Håkon Sælen, "Siblings, not triplets: social preferences for risk, inequality and time in discounting climate change" (2009) – empirical decoupling of risk, inequality and time preference: https://cccep.ac.uk/?p=17223
• K. Arrow, M. Cropper, C. Gollier, B. Groom, G. Heal, R. Newell, W. Nordhaus, R. Pindyck, W. Pizer, P. Portney, T. Sterner, R. S. J. Tol and M. Weitzman, "Determining Benefits and Costs for Future Generations", Science 341, no. 6144, 349–350 (26 July 2013), DOI 10.1126/science.1235665 – metadata and summary verified via Crossref, full text inaccessible: https://api.crossref.org/works/10.1126/science.1235665
• Moritz A. Drupp, Mark C. Freeman, Ben Groom and Frikk Nesje, "Discounting Disentangled", American Economic Journal: Economic Policy 10, no. 4, 109–134 (November 2018), DOI 10.1257/pol.20160240: https://www.aeaweb.org/articles?id=10.1257/pol.20160240
• Kevin Rennert et al., "Comprehensive evidence implies a higher social cost of CO2", Nature 610, no. 7933, 687–692 (online 1 September 2022, issue of 27 October 2022), DOI 10.1038/s41586-022-05224-9: https://www.nature.com/articles/s41586-022-05224-9
• US Environmental Protection Agency, "Report on the Social Cost of Greenhouse Gases: Estimates Incorporating Recent Scientific Advances" (November 2023), Docket EPA-HQ-OAR-2021-0317, Table ES.1: https://www.epa.gov/system/files/documents/2023-12/epa_scghg_2023_report_final.pdf
• Astrid Matthey and Björn Bünger, "Methodenkonvention 3.1 zur Ermittlung von Umweltkosten – Kostensätze", Umweltbundesamt, status 12/2020: https://www.umweltbundesamt.de/system/files/medien/1410/publikationen/2020-12-21_methodenkonvention_3_1_kostensaetze.pdf
• Astrid Matthey, Björn Bünger and Nadia Eser, "Methodological Convention 3.2 for the Assessment of Environmental Costs – Value Factors", German Environment Agency, version 10/2024: https://www.umweltbundesamt.de/system/files/medien/479/publikationen/methodological_convention_3_2_value_factors_bf.pdf
• Bundesverfassungsgericht, Order of the First Senate of 24 March 2021, 1 BvR 2656/18, 1 BvR 78/20, 1 BvR 96/20, 1 BvR 288/20 (climate protection): https://www.bundesverfassungsgericht.de/SharedDocs/Entscheidungen/DE/2021/03/rs20210324_1bvr265618.html and press release no. 31/2021 of 29 April 2021 (English): https://www.bundesverfassungsgericht.de/SharedDocs/Pressemitteilungen/EN/2021/bvg21-031.html

← All articles