Now that carbon pricing has achieved fad status in the policy world, it is time to ask how we go about estimating the right price. To answer this question, economists have long turned to Integrated Assessment Models, or IAMs. The first IAM was developed by Yale University economist William Nordhaus in the early 1990s. It consisted of a simplified representation of the economy, including energy use and CO2 emissions, coupled with a very simplified representation of the climate system, and it ran the two together so that the future effects of CO2 emissions at each point in time could be added up and discounted back to the moment of emission, yielding a time path of the optimal emission charge, or the “social cost of carbon” (SCC) as it is now called.
Where it really counts we’re using guesstimates based on inconsistent models
There are now three main IAMs in use for research and policy-making purposes. They go by the acronyms DICE, FUND and PAGE. They still all rely on simplified representations of the economy and the climate. And, most importantly, they all depend heavily on a handful of guesstimates on key parameters.
While you may not have heard of IAMs, DICE, FUND, PAGE or SCC, they now have a large and growing influence over your life, through the work of yet another acronym: IWG, which stands for the Inter-Agency Working Group of the U.S. government. The Obama administration asked the IWG to come up with an SCC estimate for use in regulation. In 2010 the IWG obtained copies of all three IAMs and ran them in-house to produce a table of numbers spanning 2010 to 2050 for discount rates from 2.5 per cent to five per cent. Using a three per cent discount rate, they reported an average SCC as of 2015 of US$26.30, rising to US$44.90 in 2050, meaning that henceforth the U.S. government could justify assigning a social cost in that range to all economic activity yielding CO2 emissions.
In 2013, the IWG updated the models and re-ran their analysis. The 2015 value at a three per cent discount rate jumped to US$38, rising to US$71 by 2050. What changed?
The equations that connect CO2 emissions to temperature change depend on a parameter called equilibrium climate sensitivity (sorry, another acronym: ECS), which is the amount of warming in degrees Celsius from doubling the amount of CO2 in the air, after the atmosphere and oceans have fully adjusted. The equations that connect temperature change to economic impacts make up what is called the damage function. The IWG made updates to the damage functions that boosted the costs, but it did not change the ECS. This is unfortunate because ECS have dropped a lot in recent years.
The IWG notes that ECS is a very uncertain concept. It could be close to zero, or as high as 10 degrees. They represent the uncertainty by using a probability distribution published in 2007 by climate scientists Gerard Roe and Marcia Baker. The IWG used this distribution to pump out thousands of possible ECS values between zero and 10 degrees, computed the SCC implied by each one, and then took the average.
But as Roe himself has pointed out, the IWG is not using this distribution correctly because it is ignoring the change in time scale associated with different ECS values. ECS is closely related to the rate of ocean heat absorption, which affects the speed of adjustment. The higher the ECS, the longer it takes the climate to adjust to higher greenhouse gas levels. Under a high-ECS case the damages occur much farther in the future and need to be discounted more heavily. But the IWG doesn’t take this into account; instead it allows high-ECS and low-ECS scenarios to occur on the same time scales, biasing the SCC upwards. To fix this, the distribution of ECS values has to be constrained to be consistent with observed ocean-heat uptake rates.
Another problem is that the IWG ignored a large and growing literature since 2012 on empirical estimates of ECS. The IWG tuned its IAMs to the Roe-Baker distribution, which matches the behaviour of climate models. In other words, economic models are calibrated to follow climate models, not real world data. It’s models all the way down.
There is now a large empirical literature on ECS estimation. About a dozen studies have appeared in the climate literature over the past half-decade using newly developed long-term data sets, and consistently the results are well below the ECS values in climate models. Despite its claim that it wanted to use “the best available science” and keep up with a “rapidly evolving field,” the IWG ignored this evidence in its entirety.
To rectify this, I teamed up with statistician Kevin Dayaratna and economist David Kreutzer at the Heritage Foundation in Washington D.C. We obtained the computer code used by the IWG for the DICE and FUND models (the PAGE code is not available for independent use) then re-ran the calculations replacing the model-based ECS values with an up-to-date empirical ECS distribution from a 2015 paper published by scientists Judith Curry and Nicolas Lewis, which constrains the ECS value using data on long-term ocean heat changes.
The result was that the SCC estimates should have been revised down, not up: indeed, a long way down. SCC estimates from the DICE model fall by 30 to 50 per cent, and from the FUND model by over 80 per cent. In fact, the FUND model yields about a 40 per cent chance that the social cost of carbon is negative, meaning that CO2 emissions are a global net benefit even before we take into account the economic benefits of the fuel consumption that generates them. FUND differs from the other two IAMs because it takes into account CO2 fertilization of plants and the observation that moderate warming in some regions will be a net benefit. The other models assume that all CO2 is bad and any temperature change (up or down) is bad, which are themselves assumptions worthy of being challenged.
To emphasize, we only changed one thing: we replaced a guesstimate of the ECS parameter with a recent estimate based on long-term observational data. And the results were dramatic. The SCC falls, and in the most detailed of the economic models, it is no longer clear that CO2 is even a net social cost.
The numbers produced by the IWG have a large and growing influence over energy and economic policy in the U.S. and Canada and elsewhere. Unfortunately, for all its claims about following the science, where it really counts it ended up peddling guesstimates based on inconsistent models. To borrow a phrase, it is time to restore science to its rightful place. Calculations behind the social cost of carbon need to reflect empirical evidence about low climate sensitivity, and when this is done, the numbers appear to be much lower than those currently in use.
Ross McKitrick is professor of economics at the University of Guelph and research chair in energy, ecology and prosperity at the Frontier Centre for Public Policy.
