A key question which has been asked about climate change for many years now is how much money should be spent today by governments on this important issue. This is a particularly critical question during the current period of high inflation and increased interest rates that are together squeezing government budgets. Every dollar spent on climate action is a dollar not spent on education, health, welfare programs, and a host of other areas which have an immediate impact.
Cost benefit analysis is a common modern tool which gets applied to deliver a supposedly objective answer to this question by weighing up the cost of climate action against the expected costs if climate change is not addressed.
In this article I’m going to argue that:
(1) the dollar value of climate impacts are going to be underestimated for the foreseeable future.
(2) reducing climate impacts to a dollar value renders some kinds of harm invisible, hiding critical moral and political questions about what we value.
(3) directly prioritising carbon reduction measures based on cost and supporting as many as necessary to reach explicit targets is likely to be much more effective.
(4) a carbon tax is only one part of the required policy mix to drive climate action.
Why is this interesting? I think that (1) probably isn’t surprising in the abstract but I still found the details of how much is missing from SCC estimates to be remarkably large. (2) is an example of a more general problem of how quantification can render invisible anything that does not get quantified - I think this should be generally interesting to EAs. (3) presents an alternative to the SCC which I hadn’t heard of before I wrote this article. (4) presents some of the latest evidence about when carbon taxes are effective and situations where other policies are a much better idea - these are details which I’d also never come across so clearly before writing this article.
(1) Underestimating The Costs Of Climate Change
The Social Cost of Carbon (SCC) is an estimate of the financial cost of each extra tonne of CO2 which is emitted into the atmosphere based on the harms that this CO2 will cause.
This simple statement hides the fact that computing a SCC value involves predicting the cost of changes to a huge number of different complex global and regional systems over the next several decades as the Earth moves into a climatic range which has never before been observed by humanity. We can only model these impacts and have very little ability to validate the predictions and rarely have the ability to account for compounding impacts across areas. Areas where we can’t model the impact are usually treated as areas that have zero impact and so the overall SCC estimate is usually very conservative.
One of the most recent attempts to compute an updated SCC in a government policy context was the US Environment Protection Agency (EPA) who published a 137 page pdf considering the latest state of SCC research as of September 2022. All the way down on page 79 of the PDF is this incredible summary table which shows which kinds of damage have been accounted for by the SCC estimates produced in the report:
For example, there’s no accounting for clearly important areas like temperature extremes or variability, or indeed changes to precipitation (rainfall), impact of coastal storms, or even displacement and migration from all of this.
Based only on the kinds of damage modelled so far, the EPA’s proposed conservative 2030 SCC is $140 (2.5% discount rate), $230 (2% discount rate), $380 (1.5% discount rate) (all in 2020 $). However, in their own words from their report:
“The modeling implemented in this report reflects conservative methodological choices, and, given both these choices and the numerous categories of damages that are not currently quantified and other model limitations, the resulting SC-GHG estimates likely underestimate the marginal damages from GHG pollution.”
(2) Hidden Moral And Political Choices
Estimating the SCC sounds superficially like an important step in working out how much money should be spent on climate action. However, any SCC estimate will necessarily obscure a large number of moral and political choices that deserve to be considered more directly. I’ll consider three examples here.
Climate change will increase the frequency of extreme heatwaves. In July 2023 Phoenix, Arizona experienced a record breaking 31 days of >110F (>43.3C) temperatures. These conditions can be partially adapted to by the use of air conditioning but what’s the right dollar price to put on this intervention? Is it the price of the air conditioning units plus the electricity it takes to run them? Should a price be put on the lived experience of sheltering indoors in air conditioned rooms day after day from daytime temperatures, or having to be continuously vigilant to avoid contact burns when you do venture outside?
Some companies are responding by designing jackets which have built in fans to help people to be able to go outside on very hot summer days. If these jackets can be made cheaply enough does that mean that higher temperature days are okay? What value should be placed on the loss of being able to go outside on a hot summer day without a fan jacket?
Climate change is pushing areas of Earth into extended droughts that are resulting in crop failures and large numbers of climate refugees. For example in 2023 the current drought in the Horn of Africa has resulted in >4 million people in need of humanitarian assistance, and at least 180,000 refugees. What price should be placed on this tragedy? Is it the cost of the humanitarian aid being provided to enable the affected people to survive? Or is there a higher price which should be placed on the collective displacement and suffering that is being directly caused by climate change?
Coral reefs are iconic sites of rich natural beauty that host more than a quarter of all marine fish species, and also directly support over 500 million people worldwide, who rely on them for daily subsistence, mostly in poor countries. In 2018 the IPCC predicted that 99% of coral reefs would be lost if the world warmed by 2C. A new study in 2022 which used a higher resolution model predicted that 99% of coral reefs may be lost much sooner when we reach only 1.5C of warming.
Attempts have been made to put a price tag on the value of coral reefs by looking at different factors such as fishing, tourism, and coastal protection. These estimates tend to be at least tens of billions of dollars per year of value. However, this again raises the question of whether any price would actually be appropriate for these rich and unique concentrations of biodiversity.
One way of addressing these issues would be to reject the requirement to reduce all the benefits down to a single dollar value which obscures that different kinds of value are being mixed together and where types of value which are hard to put a dollar price on are rendered invisible. The alternative would be to embrace a collection of measurements of the value of climate action. For example the single social cost of carbon could be replaced by an estimate of the multiple types of costs seen at a particular temperature e.g. limiting the temperature increase to 1.5C might result in $XB of financial impacts, as well as producing expected cumulative heatwave days of Y days, an expected drought area of Z km^2, and destroying A% of coral reefs.
Estimating these figures would still remain very difficult, and the choice of which measures to include would be politically contentious, but this would at least make visible some of the choices which are currently hidden while also embracing a much richer definition of harm that doesn’t have to filter harms through a financial lens that is likely to alter what is counted as important.
The philosopher C. Thi Nguyen gave a talk in 2022 titled ‘Value Collapse’ which discussed the idea that quantitative values actively make it harder to notice what gets excluded by those values. For example cost benefit analysis often end up excluding non-economic values which are then rendered invisible and of zero value. This concept should be of general interest to the Effective Altruist community as there is presumably a risk that quantitative estimates of the good done by different organisations risks rendering invisible any types of benefit which are not captured by the chosen measurement. Capitalism, power and epistemology: a critique of EA is an example of a critique of EA itself that is based on the idea that the way EA approaches quantification through a capitalist lens will shape the interventions which EA supports.
(3) A Target-Consistent Approach
The SCC is likely to be underestimated for many years to come and this implies that the costs which governments will be willing to bear will be lower than if we were actually able to fully predict the future costs of climate change.
Another key problem with a SCC based approach to determining the level of climate action is that it doesn’t directly attempt to meet specific temperature targets. This means that a SCC (and indeed any carbon tax policy where the price is determined based on anything other than a target outcome temperature) risks giving the impression of political action without actually achieving the required speed of emissions reductions needed to limit warming to a specific temperature (e.g. 1.5C).
A better approach is to take as a starting point the existing international targets which countries have already signed up to - net zero by 2050 with the Paris goal to hold “the increase in the global average temperature to well below 2°C above pre-industrial levels” and pursue efforts “to limit the temperature increase to 1.5°C above pre-industrial levels.”
The “target-consistent approach” uses a Marginal Abatement Cost (MAC) curve to set the required carbon price. This completely skips the difficult problem of computing the SCC and instead focuses on the simpler (but still complex) question of estimating the cost and potential scale of available carbon reduction interventions required to hit a particular target.
(Illustrative MAC curve - the height of each bar is the £/tCO2e of each intervention, and the width represents the scale of the available intervention. A* is the point where the cumulative emissions savings of interventions adds up to the emissions savings required to meet the target at the lowest possible price.)
The UK moved to this “target-consistent approach” in 2009 and in September 2021 released an updated carbon price based on this methodology, along with a really interesting literature review.
MACs are not a silver bullet as there are still difficult issues to consider including things like dependencies between different interventions and changes to cost as the scale of deployment of an intervention changes. “Marginal Abatement Cost Curves for U.S. Net-Zero Energy Systems” (2021) is an example of a report which attempts to address some of these issues.
The UK’s 2030 carbon value computed using a target-consistent approach is £280 (in 2020£). “To capture the full range of uncertainty, a plus or minus 50% sensitivity range has been deemed appropriate around the central series.” which gives a range of £140 - £420.
(£140, £280, £420) = ($182, $364, $546) using $1.3 to £1, compared to the EPA’s 2020 proposal of a 2030 SCC of $140 (2.5% discount rate), $230 (2% discount rate), $380 (1.5% discount rate). Using a 2022 PPP exchange rate of $1.47 = £1, this gives (£140, £280, £420) = ($206, $411.6, $617.4). This shows that the current EPA 2030 SCC estimate is within the PPP adjusted bounds of the UK’s “target-consistent approach” carbon value. However, it’s worth noting that the SCC approach has previously resulted in much lower values - in the US the 2030 SCC was first estimated to be $50 (2007 $) in 2010. The Trump administration in 2017 announced a new SCC of less than $10 based on a dramatically more conservative definition of the SCC.
Another interesting perspective is that climate action has become rapidly more affordable. This paragraph takes the history of UK action as an example. In 2003 the UK targeted a 60% reduction in CO2 emissions by 2050 and expected this to cost 0.5 - 2.0% of GDP in 2050. In 2008 the target was increased to an 80% reduction and the costs were still expected to be only 1-2% of GDP in 2050. In 2019 the target was increased to a 100% reduction with the same expected costs. In 2020 the expected net costs were estimated to have dropped to below 1% of GDP every year through to 2050. In 2021 the UK’s Office For Budget Responsibility (OBR) estimated that climate investment costs will likely be fully repaid by 2070. The OBR calculations did not include the benefits from avoided climate change impacts or other co-benefits.
Of course, agreeing the right carbon price for policy analysis is only the first step of the solution. As of June 2023 the UK government is not on track to actually deliver on its committed climate goals.
(4) Climate Policy: Beyond A Carbon Tax
One of the central policies which has been promoted for addressing climate change is a carbon tax. This is presented as a politically neutral way of pricing in the externality of carbon pollution and in theory drive decarbonisation.
The trouble is that it’s very politically challenging to introduce such a far reaching new tax, particularly at a price high enough to actually drive changes, and without causing regressive impacts.
As of 2023, 23% of global emissions are covered by some kind of carbon tax system, and according to one study:
“We conclude that the effectiveness of carbon pricing in stimulating innovation and zero-carbon investment remains a theoretical argument. So far, there is no empirical evidence of its effectiveness in promoting the technological change necessary for full decarbonization.”
Of course, you should Beware The Man Of One Study, and indeed there’s a direct rebuttal in another paper which concludes:
“A recent review study by Lilliestam, Patt and Bersalli (2021) of the impact of carbon pricing on low-carbon innovation and deep carbonization concludes that there is no evidence for such an impact. We evaluate this study and identify various shortcomings of it, which together cast strong doubts on its main conclusion. Instead, we conclude, based on the studies reviewed by the authors and additional, overlooked literature, that carbon pricing has had a small but positive and significant effect on low-carbon innovation.”
A 2018 paper provides some really tangible examples of why carbon pricing is not a panacea.
Directly subsidising a new technology can be much cheaper than taxing an existing very widely deployed technology:
“When Germany started supporting photovoltaics (PV), for example, the price differential between PV and coal was roughly US$0.50/kWh, and has since declined to essentially zero as a result of the learning triggered by the global PV expansion. Covering a large initial price differential with targeted technology support was feasible, because the initial volume of capital stock to which that support applied was so low; the total volume of financial support was still trivial in relation to the overall economy. A carbon price, by contrast, covers the price differential by penalizing the old technology, for which the capital stock starts high and then declines. Covering the $0.50/kWh price differential for PV would have required an initial carbon price of roughly $700 per ton of CO2, applied to fossil-based generators that then supplied more than 90% of the power market, and possibly to other sectors such as steel and cement. This would have crippled German industry and dramatically raised consumer prices. It was, rightly, a political non-starter.”
Planning and delivering changes to overall systems is critical to accommodate new technology:
“The second barrier is a mismatch between the new technology and the existing infrastructure. For example, it has been possible to generate small amounts of fluctuating PV and wind power without threatening the stability of power transmission and distribution grids. Today, about 15% of the European electricity supply is fluctuating PV and wind power, and it is becoming apparent that major elements of the grid, from transformer stations to long-distance transmission lines, will need to be upgraded or replaced. We will also need new infrastructures, such as large-scale electricity storage. Unless we adapt the infrastructure, the problems will only grow. Infrastructure development takes coordinated planning and development based on long-term strategic priorities, and this is not something that carbon prices directly address.”
Institutions and markets may need to be reformed:
“The third barrier is institutional. One example is to be found in the rules determining wholesale power prices. Wind power and PV generation respond to the weather rather than to price signals. Under current power market designs, growing shares of wind power and PV have pushed down wholesale power prices, in some cases well below zero, precisely during their times of peak production. This reduces profits for fossil generation, but even more so for wind and solar themselves. Carbon prices do not directly address this problem, whereas market reforms can.”
I found a great 2021 discussion paper which reviews the controversy around carbon prices. This paper proposes a hybrid of three different approaches:
(1) “Especially in the middle of the marginal abatement cost curve, carbon pricing may induce emission reductions, for instance by triggering fuel-switching in the power sector. The efficiency of carbon pricing to achieve emission reductions at the margin through process optimization has been amply demonstrated. Assuming a certain degree of rationality and market functioning, carbon pricing initiatives focus mitigation efforts on activities with the lowest abatement cost, and are therefore an important element in any climate policy portfolio to drive emissions reductions cost-effectively in the near term.”
(2) “Behavioral economics provides a theoretical explanation of barriers to energy efficiency, which are difficult to address with carbon pricing. On the left side of the marginal abatement cost curve, where financial incentives tend to be ineffective, information policies such as labelling and command-and-control policies such as performance standards, phase-outs, or outright bans tend to be more effective. Bans of inefficient light bulbs are a poster example. Pricing mechanisms can provide incentives, but will not prevent people from engaging in certain conducts.”
(3) “Public choice theory explains challenges faced in making collective decisions, which, in turn, have vital consequences for policy making to foster innovation. In particular, at the right side of the marginal abatement cost curve, technologies can be found which require large investments and long-term horizons. It is important to identify such areas which actually require complementary policies in order to avoid supporting dead-end technologies while precluding valuable alternatives that are either unknown to policy-makers, or underappreciated in their mitigation potential. Directed policies sacrifice the availability of those technology options which are not favoured by support policies, whereas carbon pricing creates a technology-neutral incentive for all mitigation options.”
There are many examples of these non-carbon tax policies, for example:
- Type (2) - Phasing out sale of Internal Combustion Engine (ICE) cars, some countries in 2030, and many others by 2040 (source)
- Type (3) - $94BN for Clean Energy Technology Demonstrations, announced in Sept 2022 (source)
The hardest part of introducing a carbon price is that it increases the price of existing products. This is extremely unpopular and has lead to things like the street riots led by the ‘Yellow Vests’ movement in France in late 2018.
The alternative approach is to rely on technology innovation and learning curves to drive down the price of green options to the point where switching technology is cheaper without subsidies. The term “green premium” was coined in 2020 to refer to the difference in price between the green option and the conventional option.
We’re already seeing the green premium being eliminated in some key areas of the energy transition such as electricity generation: in 2021, IRENA reported that two-thirds of newly installed renewable power in G20 countries had lower costs than the cheapest fossil fuel-fired option. This is resulting in important developments such as India deciding to halt consideration of any proposals for new coal plants for the next five years and focus on growing its renewables sector instead.
The key observation about green premiums is that we can’t be sure how quickly (if ever) innovation will eliminate these for all the required technologies in the energy transition. The IEA estimates that in 2023 only three of more than fifty technology areas are “on track” in the energy transition. Hitting emissions reductions targets will likely involve transitioning to many technologies before their green premium has been eliminated. This will require governments to have the political will to find the money necessary to provide subsidies that (hopefully temporarily) cover the green premium, possibly through increased taxes or by cutting spending elsewhere, or regulations will have to be introduced to force consumers to buy the more expensive green options. The only other alternative is that emissions reductions will be delivered late and the world will warm more than we want it to.