Can you trust a model whose output keeps changing? Interpreting changes in the social cost of carbon produced by the DICE model

Abstract

The social cost of carbon (SCC) measures the present value of the economic damages caused by emitting one marginal ton of carbon dioxide into the atmosphere. It plays a crucial role in climate policy analysis, where it is used to suggest optimal carbon prices or quantify the benefits of actions that reduce emissions. One prominent framework used to estimate the SCC is the Dynamic Integrated Climate-Economy (DICE) model. As updated versions of DICE have been introduced, its SCC estimates have changed, sometimes by amounts that would appear significant. For example, the SCC in 2020 produced by DICE rose 54% from its 2013R version to its 2016R2 version. We address two important questions. First, what changes to DICE explain this increase in its SCC? Second, how surprising is the magnitude of this increase, relative to the uncertainty present in DICE’s input parameters? We find that changes in scientific parameters and updated initial conditions due to near-term forecasting errors accounted for the largest shares of the SCC increase. The later SCC estimate falls within the 80% probability interval produced using the earlier model with uncertainty. Therefore, the 54% increase should not be considered surprising or dispositive regarding the quality of DICE itself.