Maximilian Auffhammer at the Energy Institute at Haas reviews a new paper that suggests CO2 causes over three times as much damage in dollar terms as the figure currently used by the US government, $51 per ton. The new study shows $185 per ton of CO2 as the Social Cost of Carbon (SCC). The updated model is superior to previous models, says Auffhammer. It’s also open source, so anyone can use it, criticise it, and tweak the numbers to get different results depending on their own assumptions. Auffhammer describes it as a major step forward in SCC modelling, but says it can still be improved further. Only damage to the four biggest sectors is included: agriculture, energy, sea level rise, and mortality. What about adding species loss, forests, water availability, conflict, and migration? What about “equity weighting”, to reveal how the poor are more affected by the wealthy? And there’s little measurement of the, admittedly, difficult-to-measure “fat tail risk” of a catastrophic event with a low probability of happening.
The Social Cost of Carbon
The third floor of the newly seismically retrofitted Giannini Hall at Berkeley is an awesome place to be, if you care about one of the more obscure numbers in existence: The Social Cost of Carbon. The SCC is the damage one ton of CO2 does over its long lifetime across all sectors and the entire planet.
The sheer number of nerds obsessed with how this number is calculated, permanently located, or visiting this single hallway makes it a super fun place to be – for those nerds. The level of excitement in response to the release of a new paper led by the younger German economist on this floor was significantly higher than when the Warriors deservedly won the NBA championship last year. Sorry Boston, but you know it’s true.
Cost-benefit analyses are widespread now
One of the biggest contributions (at least I think so) of economists to public policy is the required application of Benefit Cost analysis for a significant share of federal regulations since the 1970s (or 1993 depending on how you feel). If federal agencies want to impose a new regulation, often they have to show as part of a regulatory impact analysis that the benefits of the regulation (e.g. avoided pollution damages from lower energy consumption due to energy efficiency rules) are greater than the costs (higher cost of manufacturing said more efficient gadgets).
Since basically every human activity causes some greenhouse gas emissions, the damage a ton of CO2 does is key. There is a long history of using this number going back to the Bush II presidency when NHTSA, EPA, and DOE applied three very different numbers for the same gas emitted. During the Obama presidency, an interagency working group produced an official number, which was then $42 per ton emitted in 2020 using a 3% discount rate.
Let’s skip the bad economics that happened under President Trump. President Biden in the first month in office put in place a slightly updated SCC, which was $52/ton and ordered a significant update, which was supposed to take into account the suggested improvements by the National Academies of Science and Engineering.
The new paper by led by David Anthoff and Kevin Rennert presents the results of this immense amount of work conducted mostly during the Trump years with a team organised out of RFF composed of a team of future all stars (shoutout to Frank Errickson, Lisa Rennels, and Cora Kingdon) and current all stars (who do not need a shoutout).
Because we have all read the Lord of the Rings too many times, the global press focused on the number that rules them all: The new central estimate of $185 per ton of CO2 which is more than thrice the current number (at a 2% risk free rate used to discount). Extra, extra, read all about it! Bigger numbers! Yuge! RaRaRa. Sure. But here is what I love about this paper:
- It lives in the light. I mean it’s all open source. If you can program in Julia (which you can learn in a weekend if you know some Python or R) you can modify the model in any way your little heart desires. You can add stuff. Subtract stuff. Change stuff. Make it yours in any way you want. While Richard Tol will remind me that the great grandparents of the new model (DICE and FUND) also lived out in the open and that Fran Moore (currently in the White House) dragged the third (PAGE) into the light, I would like to remind us that most of the components of these older models were calibrated when Counting Crows were topping the charts. Counting who? Exactly.
- The damage functions are updated and account for some adaptation. There are four sectors in the model: agriculture, energy, sea level rise, and mortality. The damage functions, which translate a changed climate into changed welfare have been at the focus of the recent empirical economics literature and this effort draws heavily from this literature (the Climate Impact Lab at Berkeley/Chicago/Rhodium/Rutgers has a parallel effort with a different approach). Why does this matter? For some of the older models, the damages on the agricultural and energy sectors were not consistent with the most recent literature. The update changed the relative contributions of these sectors.
- The paper makes a significant update to discounting. I am not going to get into the technicalities here, but there are two things. First, the paper employs a 2% risk free rate, which is lower than the lowest rate used before and is consistent with recent peer review top journal yadda yadda expert elicitation. The second update is that discounting now takes into account the rate of economic growth, which is key if you are a Ramsey kind of person (sorry for the nerd lingo) and care about pricing risk correctly.
- The coolest thing, however, IMHO, is the full characterisation of uncertainty. Soup to Nuts. These models require many things. The first thing you need is future emissions based on income and population assumptions. In the old days, we just put together a handful of scenarios that seemed reasonable and did not really attach probabilities to earths with different levels of wealth and population. This paper went bonkers on this by combining some cool new econometrics on really long run forecasts with expert elicitation to characterise future states of the economy probabilistically! This sounds straightforward, but it’s not and this alone would have been a major step forward.
- They do not stop there. The authors update the climate model to reflect current scientific understanding, a damage component, and a discounting module. Each of these has its own source of uncertainty – some related to assumptions about the future, some related to assumptions about parameters. But this new model (GIVE) allows us to take into account the change in uncertainty and translates this into a distribution of the SCC taking these different sources of uncertainty into account. So. Flipping. Cool.
What’s still missing from the modelling?
There is a bunch of other neat stuff in here. And some Debbie or Donald Downers in the comments and on Twitter are gonna spread some hate with respect to what the authors should have done to satisfy their own priors. But. I am going to put a few things out there that we should push hard on to better characterise the social cost of carbon.
- There are literally hundreds of sectors out there possibly affected by climate change. This model contains four. Granted these are probably the most climate sensitive ones we know something about. Yet, I want us to push hard on non-market and non-market”ish” damages like species loss, forests, water availability, conflict, migration. There are some really cool folks like Hannah Druckenmiller working on these issues, but this is an all hands on deck kind of situation. Get to it.
- I still do not understand why we do not put the equity weighted Social Cost of Carbon centre stage? We keep on talking about environmental justice like we know what we are talking about and care, and then we focus on the non-equity weighted number. I have written about this before and my inbox was full of “Woah. This is cool. How do I do that?” We know that the Germans use an equity weighted SCC, which David Anthoff was involved in calculating. So I really hope that David will not get distracted by the disaster that is currently unfolding in our favorite Fussball Team and provide us with an equity weighting module.
- We have little to no handle on fat tail risks (really big catastrophic events with non-zero probabilities. Pick your poison.) I continue to think that we may want to contemplate some “Weitzman disclaimer” to any SCC we deploy in Benefit Cost Analysis. To quote one of my heroes from his superb 2009 paper: “Perhaps in the end the climate-change economist can help most by not presenting a cost-benefit estimate for what is inherently a fat-tailed situation with potentially unlimited downside exposure as if it is accurate and objective? […] Even just acknowledging more openly the incredible magnitude of the deep structural uncertainties that are involved in climate-change analysis and explaining better to policymakers that the artificial crispness conveyed by conventional IAM-based CBAs here is especially and unusually misleading compared with more ordinary non-climate change CBA situations might go a long way toward elevating the level of public discourse concerning what to do about global warming.”
This last point is what keeps me awake at night. A graduate student walked into my office and asked me “How do I model and learn about something that has not happened in the past but might happen in the future?” Our friends in climate science have literally spent billions of dollars on this question. We economists in the climate landscape have mostly focused on learning from measurable things in the rearview mirror, which is what the toolkit-du-jour is geared towards. But I worry that in a world without the great Marty Weitzman in it, we are not sufficiently nudged to encourage this graduate student and her peers to pursue these all important questions.
Maximilian Auffhammer is the George Pardee Professor of International Sustainable Development at the Energy Institute at Haas, part of the University of California, Berkeley.
This article is published with permission
Keep up with Energy Institute blogs, research, and events on Twitter @energyathaas