Name: Dynamic Integrated model of Climate and the Economy (DICE)
Type: IAM
Originator: William Nordhaus
Documentation: Documentation and model
Geographic coverage: Global
Description: DICE offers an internally consistent framework based on a standard Ramsey growth model for analyzing the interplays between the macroeconomy, greenhouse gas emissions, climate policies, and climate change. Key elements include portable modules and quantifications for climate change damage functions, dynamic estimates of aggregate emissions reduction costs, a simplified carbon cycle-climate system representation, dynamic estimates of the social cost of carbon, and a flexible discounting module. Major innovations in DICE-2023 (updated from DICE-2016) include (1) a new carbon cycle-climate system representation, (2) an updated damage function based on a synthesis of 56 estimates across 33 published studies that includes post-2016 research, (3) a new representation of non-carbon dioxide greenhouse gas emissions and abatement, and (4) a new approach to discounting that incorporates uncertainty. These updates lead to a substantially higher social cost of carbon (SCC), a lower cost of maintaining the 2°C limit on temperature rise, and a lower cost-benefit optimal emission and warming profile than in previous versions.
Questions to be answered/variables considered: The model can be used to (1) quantify the SCC, quantify costbenefit optimal climate policy paths under different parameter choices, (3) quantify cost-effective policy paths given policy targets, and (4) characterize the costs and benefits of arbitrary policy paths under different parameter scenarios. Endogenous outputs include GDP, climate change damages, mitigation expenditures, consumption/ investment, carbon prices, the SCC, industrial carbon emissions, land-use carbon emissions, abatable non-carbon dioxide emissions, global mean surface temperature change, and carbon concentrations.
MoFs can use DICE (or RICE, the multi-region version of DICE) to help inform long-run macroeconomic and fiscal projections of global or regional GDP impacts of different climate policy scenarios, and output on the SCC can inform carbon pricing policies, public cost-benefit analysis, and setting subsidy rates. Moreover, DICE-2023 model elements can be integrated into, e.g., models of short-run economic fluctuations such as DSGE models or New Keynesian frameworks, to help address questions DICE is not designed to answer.
Strengths:
- Modeling elements are simple, flexible, and portable.
- Given DICE’s simplicity, uncertainty and sensitivity analyses are relatively easy to conduct.
Limitations:
- DICE’s simplicity means it abstracts from many complexities of modern macroeconomies.
- As the time step is five years, DICE is not suited to study short-run macroeconomic frictions and fluctuations.
- The model focuses on a representative consumer and final goods production sector.
- The multi-regional version of DICE, RICE, allows country- and region-level analysis, but the model is not designed to answer some of the granular questions relevant for MoFs (e.g., on targeting clean technology subsidies or the distributional impacts of carbon pricing).
Assumptions: The evolution of technology and emissions reduction costs are taken to be exogenous, i.e., not affected by climate policy. The cost of mitigating carbon emissions is taken to be a proportional and contemporaneous fraction of GDP, which increases nonlinearly in climate policy stringency. Population growth is taken to be fixed, though the mortality impacts of climate change are valued in the damage function. Climate change damages are assumed to be quadratic in global mean surface temperature change. This is in line with estimates for modest temperature change, but evidence on damages is very limited for higher levels of warming, and a damage function does not reflect threshold damages.
Use: The code, user manual, and source data are publicly available and can be readily modified by users. DICE runs on GAMS, and thus a GAMS license and programming expertise are needed to use the model. An Excel version is available but comes with additional caveats noted in the documentation. For older versions of DICE, MATLAB code is publicly available from other scholars.