Notes on the social cost of carbon

A recent paper by Ackerman and Stanton did some re-estimates of the Social Cost of Carbon, finding this measure of the externality (or costs imposed on third parties) from burning fossil fuels could be as high as $893 per tonne of CO2, rising to $1,500 per tonne by 2050. These are extreme estimates, but they are calculated using a conservative model. The key point is that the official US government measure, that is $21 per tonne, was unrealistically low because of a range of assumptions in the model itself, as well as many other things the model does not take into account.

Some context: Frank Ackerman has long been a critic of attempts to put market value on environmental problems (correctly, in my view). I first came across his work when I did some research on cost-benefit analysis of environmental regulation. A range of questions inevitably arise about how one values human injury, disease or death in such exercises, not to mention ecosystem damage, extinct species and so forth that have no market value per se.

In recent years, Ackerman (with Stanton) has made a devastating critique of estimates of the social cost of carbon. A year ago, a paper deconstructed the concept and its particular application in a handful of statistical model used to derive the US government number, which (surprise, surprise) ends up being pretty low. SCC is anthropocentric, so there is no cost to other catastrophes and plant/animal extinctions beyond any utility value to human consumers (and discounted for future generations). It cannot deal well with events of low probability but catastrophic consequence.

The new paper follows up that critique using the conservative DICE model developed by Nordhaus to show that amending the standard and unrealistic assumptions leads to very different results, in the hundreds of dollars per tonne range with that top end number of $893. Basically, you get larger values of SCC when either damages are higher and/or when the discount rate is lower. Business as usual is basically human extinction, so damages are about as high as they could get.

Somewhere in the $150-500 range of carbon price, we get “maximum feasible abatement” — all options for reducing emissions are economical — meaning numbers over say $500 per tonne do not change the imperative for action in dollar terms, they just urge even greater precautionary action. Now. The authors in an oped make the comparison to fire insurance:

Your house might not burn down next year. So you could probably save money by cancelling your fire insurance. That’s a “bargain” that few homeowners would accept. But it’s the same deal that politicians have accepted for us, when it comes to insurance against climate change. They have rejected sensible investments in efficiency and clean energy, which would reduce carbon emissions, create green jobs, and jumpstart new technologies – because they are too expensive.

While your house might not burn down, your planet is starting to smolder. Extreme weather events are becoming more common, and more expensive: in the first half of 2011, Mississippi River floods cost us between $2 and $4 billion, while the ongoing Texas drought has cost us between $1.5 and $3 billion, according to the National Climatic Data Center. And there’s much worse to come: climate-related extremes are already forcing millions of people from their homes worldwide; ice sheets and glaciers are melting much faster than expected; the latest research shows we are rapidly heading for summer temperatures at which crop yields in America will start to plummet.

… Our report finds deep flaws in the U.S. government’s $21 per ton estimate. That inaccurate estimate promotes inaction, with enormously harmful consequences. Our research incorporates an up-to-date understanding of climate risk and uncertainty, and finds that the true cost of carbon emissions could be almost $900 per ton today, and more than $1,500 by 2050. Granted, these are the high end of the range of 16 scenarios that we studied. We aren’t sure that the costs will be that high – but we also can’t be sure that climate change won’t be that expensive. It’s the fire insurance problem: you buy insurance because you can’t be sufficiently sure that your house won’t burn down.

Just as climate models have ended up being too optimistic matched up against real-world data, so the conservative approach to modeling financial costs and SCC is showing the same flaws. The A-S paper also includes an additional commentary by Simon Dietz on the social cost of carbon that notes that under standard assumptions, the model finds a loss of GDP of 50% when temperature increase is 18 degrees C, which is clearly out of whack with science.

The Stern Review prompted a lot of debate about discount rates. He advocated for a zero discount rate, which is to say that future generations’s welfare is treated like our own. We do not “discount” the impact of bad things that happen. It means “let’s party until the house falls down” so if not precisely zero the discount rate for such exercises should be very close to zero.

There is good overview article, “Discounting and the social cost of carbon”, that lays out what happens when you use different discount rates. For example, a 3% discount rate means you value people 100 years from now at only 5% of someone today, at 5%, only at 1%; and even if we drop the discount rate to 1%, we are still only valuing someone a century onward at 37%. Anything beyond this, like 500 years, has essentially zero bearing from a present value point of view, so over the long time frames climate analysis takes place discounting is rigged towards doing very little today.