Second in a series of posts on climate policy. Find Part 1 here.
According to scientists, climate damages are deeply uncertain but could be ominously large (see the previous post). Alternatively, according to the best-known economic calculation, lifetime damages caused by emissions in 2020 will be worth $51 per metric ton of carbon dioxide, in 2018 prices.
These two views can’t both be right. This post explains where the $51 estimate comes from, why it’s not reliable, and the meaning for climate policy of the deep uncertainty about the value of damages.
A tale of three models
The “social cost of carbon” (SCC) is the value of present and future climate damages caused by a ton of carbon dioxide emissions. The Obama administration assembled an Interagency Working Group to estimate the SCC. In its final (August 2016) revision of the numbers, the most widely used variant of the SCC was $42 per metric ton of carbon dioxide emitted in 2020, expressed in 2007 dollars – equivalent to $51 in 2018 dollars. Numbers like this were used in Obama-era cost-benefit analyses of new regulations, placing a dollar value on the reduction in carbon emissions from, say, vehicle fuel efficiency standards.
To create these numbers, the Working Group averaged the results from three well-known models. These do not provide more detailed or in-depth analysis than other models. On the contrary, two of them stand out for being simpler and easier to use than other models. They are, however, the most frequently cited models in climate economics. They are famous for being famous, the Kardashians of climate models.
DICE, developed by William Nordhaus at Yale University, offers a skeletal simplicity: it represents the dynamics of the world economy, the climate, and the interactions between the two with only 19 equations. This (plus Nordhaus’ free distribution of the software) has made it by far the most widely used model, valuable for classroom teaching, for initial high-level sketches of climate impacts, and for researchers (at times including myself) who lack the funding to acquire and use more complicated models. Yet no one thinks that DICE represents the frontier of knowledge about the world economy or the environment. DICE estimates aggregate global climate damages as a quadratic function of temperature increases, rising only gradually as the world warms.
PAGE, developed by Chris Hope at Cambridge University, resembles DICE in level of complexity, and has been used in many European analyses. It is the only one of the three models to include any explicit treatment of uncertain climate risks, assuming the threat of an abrupt, mid-size economic loss (beyond the “predictable” damages) that becomes both more likely and more severe as temperatures rise. Perhaps for this reason, PAGE consistently produces the highest SCC estimates among the three models.
FUND, developed by Richard Tol and David Anthoff, is more detailed than DICE or PAGE, with separate treatment of more than a dozen damage categories. Yet the development of these damages estimates has been idiosyncratic, in some cases (such as agriculture) relying on relatively optimistic research from 20 years ago rather than more troubling, recent findings on climate impacts. Even in later versions, after many small updates, FUND still estimates that many of its damage categories are too small to matter; in some FUND scenarios, the largest cost of warming is the increased expenditure on air conditioning.
Much has been written about what’s wrong with relying on these three models. The definitive critique is the National Academy of Sciences study, which reviews the shortcomings of the three models in detail and suggests ways to build a better model for estimating the SCC. (Released just days before the Trump inauguration, the study was doomed to be ignored.)
Expected climate damages are uncertain over a wide range, including the possibility of disastrously large impacts. The SCC is a monetary valuation of expected damages per ton of carbon dioxide. Therefore, SCC values should be uncertain over a wide range, including the possibility of disastrously high values.
Look beyond the three-model calculation, and the range of possible SCC values is extremely wide, including very high upper bounds. Many studies have adopted DICE or another model as a base, then demonstrated that minor, reasonable changes in assumptions lead to huge changes in the SCC. To cite a few examples:
- A meta-analysis of SCC values found that, in order to reflect major climate risks, the SCC needs to be at least $125.
- A study by Simon Dietz and Nicholas Stern found a range of optimal carbon prices (i.e. SCC values), depending on key climate uncertainties, ranging from $45 to $160 for emissions in 2025, and from $111 to $394 for emissions in 2055 (in 2018 dollars per ton of carbon dioxide).
- In my own research, coauthored with Liz Stanton, we found that a few major uncertainties lead to an extremely wide range of possible SCC values, from $34 to $1,079 for emissions in 2010, and from $77 to $1,875 for 2050 emissions (again converted to 2018 dollars).
- Martin Weitzman has written several articles emphasizing that the SCC depends heavily on the unknown shape of the damage function – that is, the details of the assumed relationship between rising temperatures and rising damages. His “Dismal Theorem” article argues that the marginal value of reducing emissions – the SCC – is literally infinite, since catastrophes that would cause human extinction remain too plausible to ignore (although they are not the most likely outcomes).
Whether or not the SCC is infinite, many researchers have found that it is uncertain, with the broad range of plausible values including dangerously high estimates. This is the economic reflection of scientific uncertainty about the timing and extent of climate damages.
How much can we afford?
As explained in the previous post in this series, deep uncertainty about the magnitude and timing of risks stymies the use of cost-benefit analysis for climate policy. Rather, policy should be set in an insurance-like framework, focused on credible worst-case losses rather than most likely outcomes. Given the magnitude of the global problem, this means “self-insurance” – investing in measures that make worst cases less likely.
How much does climate “self-insurance” – greenhouse gas emission reduction – cost? Several early (2008 to 2010) studies of rapid decarbonization, pushing the envelope of what was technically feasible at the time, came up with mid-century carbon prices of roughly $150 – $500 per ton of carbon dioxide abated. Since then, renewable energy has experienced rapid progress and declining prices, undoubtedly lowering the carbon price on a maximum feasible reduction scenario.
Even at $150 to $500 per ton, the cost of abatement was comparable to or lower than many of the worst-case estimates of the SCC, or climate damages per ton. In short, we already know that doing everything on the least-cost emission reduction path will cost less, per ton of carbon dioxide, than worst-case climate damages.
That’s it: end of economic story about evaluating climate policy. We don’t need more exact, accurate SCC estimates; they will not be forthcoming in time to shape policy, due to the uncertainties involved. Since estimated worst-case damages are rising over time, while abatement costs (such as the costs of renewables) are falling, the balance is tipping farther and farther toward “do everything you can, now.” That was already the correct answer some years ago, and only becomes more correct over time.
That’s not the end of this series of blog posts, however. Three more are coming, addressing three policy problems that arise in climate advocacy: how to talk about methane and natural gas; taxes versus cap and trade systems; and the role of equity and economic obstacles to climate policy.