Policies, Projections, and the Social Cost of Carbon: Results from the DICE-2023 Model

Abstract

The present study examines the assumptions, modeling structure, and preliminary results of DICE-2023, the revised Dynamic Integrated Model of Climate and the Economy (DICE), updated to 2023. The revision contains major changes in the carbon and climate modules, the treatment of non-industrial greenhouse gases, discount rates, as well as updates on all the major components. The major changes are a significantly lower level of temperature of the cost-benefit optimal policy, a lower cost of reaching the 2° C target, an analysis of the impact of the Paris Accord, and a major increase in the estimated social cost of carbon.

...

Table 7 and Figure 7 show estimates of the social cost of carbon (SCC). The SCC in the baseline run is $61/tCO2 for the 2020 period (in 2019 international $). This is above the SCC for the C/B (Cost/Benefit) optimal run of $53/tCO2 because damages are smaller in the C/B optimum. It is far below the SCC for the 2 °C run of $85/tCO2. The higher SCC in the temperature-limited run reflects the economic interpretation that a tight temperature limit is equivalent to a damage function with a sharp kink at the temperature limit and therefore to a sharply higher damage function above 2 °C. Note that the estimates of the SCC in the current DICE version are significantly above those in earlier vintages for reasons discussed in other sections, see particularly the next section. 

One of the most instructive findings involves the importance of discounting for the SCC and other policies. Table 7 shows the powerful impact of discounting on the SCC. The social cost of carbon at a 5% discount rate is two-thirds of the DICE C/B optimal estimate for 2020, while that of a 1% discount rate is 8 times the DICE C/B optimal estimate for 2020.

Additionally, Figure 8 compares estimates of the SCC with several other current values. The GIVE model is a comprehensive estimate prepared by researchers at Resources for the Future using probabilistic estimates of output and other components of damage estimates (Rennert et al., 2022). It uses a relatively low discount rate and has a relatively high social cost of carbon. A second set of estimates pertains to the SCC used by the federal government and prepared by an interagency working group. Figure 8 shows draft SCC estimates from EPA (2022) for both their overall assessment and specific to a damage module based on the DSCIM model (Climate Impacts Lab, 2022) for near-term discount rates from 1.5% to 2.5%. Conditional on discounting assumptions, the EPA estimates align very closely with those of DICE-2023. Figure 8 also shows a draft update (OMB, 2021) based on earlier methods and models which did not contain recommended methodological updates. This estimate is notably lower than the corresponding value in DICE-2023. The key takeaway from Figure 8 is the importance of the discount rate in determining the SCC.

A major change in the results of the DICE model over the years has been the rising estimates of the social cost of carbon. The original DICE-1992 model did not calculate a SCC, which came later to climate-change economics. However, rerunning the baseline scenario for the 1992 model gives an estimate of $18/tCO2 compared to $61/tCO2 in the 2023 model (in 2019$). The upward revision is a notable illustration of the evolving scientific understanding of damages, discount rates, and levels of output. Further research will provide a decomposition of the sources of the change in SCC due to different components.

by Lint Barrage & William D. Nordhaus

National Bureau of Economic Research (NBER) www.NBER.org

Working Paper 31112; Issue Date: April, 2023

https://www.nber.org/papers/w31112