Four witnesses testified on “the Obama administration’s Social Cost of Carbon” at a House Resources Committee hearing last week. Today’s post highlights key points from the testimonies of Cato Institute scientist Patrick Michaels and Heritage Foundation economist Kevin Dayaratna.
First some background. The social cost of carbon (SCC) is the present discounted value of cumulative damages allegedly inflicted on society by an incremental metric ton of carbon dioxide (CO2) emissions over a period of decades to centuries.
Discernible in neither meteorological nor economic data, carbon’s social cost exists in the virtual world of “integrated assessment models” (IAMs) — computer programs that combine speculative climatology with speculative economics. By fidding with non-validated climate parameters, made-up damage functions, and below-market discount rates, SCC analysts can get almost any result they desire.
What they typically desire is to make fossil fuels look unaffordable no matter how cheap, and renewables look like a bargain at any price. However curious as an academic exercise, when used to make or influence public policy, SCC analysis is computer-aided sophistry.
The Obama administration’s Interagency Working Group (IWG) on the social cost of carbon uses three IAMs — DICE, FUND, and PAGE — to estimate SCC values. EPA and the Department of Energy routinely incorporate SCC estimates into the cost-benefit analyses they use to justify their regulatory proposals. The White House now requires all federal agencies to incorporate SCC estimates in environmental impact reviews under the National Environmental Policy Act (NEPA).
Agencies have an incentive to periodically increase SCC estimates to make their regulations look more beneficial. For example, the IWG’s 2013 technical support document (TSD) increased the SCC values of the group’s 2010 TSD by roughly 60%.
In other words, in just four short years, climate change somehow got 60% worse and CO2 reductions 60% more valuable. Yet during that period, climate models increasingly overestimated global warming, and multiple datasets still showed no clear link between climate change and the frequency and strength of storms, droughts, and floods. Your tax dollars at work!
Earlier this month, the IWG published its response to comments (RTC) on the 2013 TSD. The IWG made minor technical corrections in how it runs the DICE and PAGE models but did not accept any of the substantive corrections recommended in 150 significant comment letters.
With that as context, let’s turn to the testimonies.
A critical input in SCC calculations is climate sensitivity, an estimate of how much warming results from a doubling of CO2-equivalent greenhouse gas concentrations. Both the 2013 TSD and the IWG’s July 2015 RTC reaffirm the IPCC 2007 Fourth Assessment Report’s (AR4) “best” climate sensitivity estimate of 3ºC.
That is unacceptible, Michaels argues. Some 14 recent studies and 20 experiments estimate lower sensitivities, averaging about 2ºC.
Figure explanation: The black arrow indicates the 90% confidence interval of the IWG’s sensitivity estimates, based on the probability distribution in Roe and Baker (2007). The black vertical line indicates the IWG’s median sensitivity estimate. The colored arrows indicate the 5% to 95% confidence intervals of recent sensitivity determinations. Colored vertical lines indicate the median or best estimates in recent studies. Spencer and Braswell (2013) present a single sensitivity estimate based on ocean heat content data.
Using updated sensitivity estimates and probability distributions from recent literature would reduce IAM sensitivity estimates by 35%-60%.
Of particular importance, updating IAM sensitivity assumptions would significantly reduce the probability of high-end (≥4.5ºC) warming scenarios and the associated risk of catastrophic events such as collapse of the Greenland ice sheet. Michaels faults the IWG for refusing to recognize “what is now becoming mainstream science.”
Worse, the administration’s SCC analysis is structurally biased. Literally thousands of empirical observations show that rising CO2 concentrations increase crop yields. Combining such information with economic data from the Food and Agriculture Organization on 45 of the world’s most important food crops, climate researcher Craig Idso estimates that CO2 fertilization boosted global agricultural output by $3.2 trillion during 1961-2011 and will increase output by another $9.8 trillion during 2012-2050. Yet, Michaels observes, “only one of the three IAMs used by the IWG has any substantial impact from carbon dioxide fertilization, and the one that does [FUND], underestimates the effect by approximately 2-3 times.”
If all three models incorporated Idso’s CO2 fertilization effect, SCC estimates would be very low or, in some computations, even “negative,” meaning the net impact is beneficial.
In any event, models lacking significant CO2 fertilization — a “known physical effect” — should be disqualified from use in regulatory analysis.
Dayaratna, who has run the DICE and FUND models with reasonable alternative inputs, finds that IAMs are “far too sensitive to the modeler’s assumptions to be legitimate tools for regulatory policy.” Dayaratna and his colleagues examined how assumptions regarding three key inputs — discount rates, time horizon, and climate sensitivity distribution — determine SCC estimates.
Climate change mitigation is an investment with relatively predictable near-term costs and largely (or entirely) speculative payoffs decades and even centuries into the future. Whether mitigation is a prudent investment of scarce resources critically depends on the discount rates selected to estimate the present value of avoided CO2 emissions in 2050, 2100, and beyond.
OMB Circular A-4 instructs agencies to use discount rates of both 7% (the average long-term rate of return on private capital investment in the U.S. economy) and 3% (the average rate of return on long-term government bonds) when estimating the costs and benefits of regulation. However, both the 2010 and 2013 TSDs declined to calculate SCC values using a 7% discount rate.
Small wonder. When Dayaratna and his colleagues ran the DICE and FUND models with a 7% discount rate, SCC estimates declined “drastically.” Some estimates in the FUND model even became negative.
The damage functions in IAMs — equations specifying how increases in average global temperature affect global consumption and GDP — are partly based on assumptions about how adaptive capabilities will evolve as the world warms. Adaptive capability chiefly depends on technology. But nothing is harder to predict than long-term technological change, and nobody has a clue what the economy of 2300 will look like.
So Dayaratna ran the DICE model with a slightly less unrealistic time horizon of 150 years. Again, SCC estimates drop substantially, in some cases as much 25%.
Corroborating Michaels’s testimony, Dayaratna finds that when DICE and FUND are run with updated climate sensitivity distributions — Otto et al. (2013) and Lewis (2013) — warming projections and the associated SCC values drop dramatically.
Dayaratna finds that when the FUND model is run with just two of the three reasonable alternative inputs — updated sensitivity estimates and a 7% discount rate — “the SCC is overwhelmingly likely to be negative, which would suggest the government should, in fact, subsidize (not limit) carbon dioxide emissions.”
No doubt the probability of negative SCC values would be even higher if, in addition, the FUND model were run with Idso’s CO2 fertilization benefit.
Dayaratna concludes that even if IAM damage functions were not “essentially arbitrary,” SCC calculation models are “far too unreliable for use in energy policy rulemaking.”