Methodological and empirical flaws in the design and application of simple climate-economy models

Abstract

One of the main arguments brought forward in favour of the continued use of simple climate-economy models is their transparency, which should enable researchers to easily interpret the simulation results and adapt the model to their specific research interests. We investigate the degree to which this claim is supported in the case of the DICE model but most of our findings are relevant for other welfare-optimizing climate-economy models as well. Specifically, this paper reviews the handling of time discounting in social welfare functions, the combination of different social welfare functions in an analysis, the calibration of uncertain climate parameters, the representation of uncertainty about future climate change, and the evolution of carbon abatement costs over time. We find that each of these aspects has been treated inconsistently in the past, and that these inconsistencies can strongly affect the results of several previous studies. We discuss the methodological questions raised by some of these problems and make specific recommendations how to avoid the problems identified here in future analyses.

How well do integrated assessment models simulate climate change?

“…A qualitative comparison has been made by Goodess et al (2003), while some papers have looked at the climate-carbon cycle behaviour of individual IAMs. Schultz and Kasting (1997), Joos et al (1999) and Füssel (2007), for instance, evaluated the behaviour of the DICE-1999 model with respect to simulating climate and the carbon cycle. Individual SCMs developers discuss the quality of their calibration in comparison to more sophisticated models as well (e.g.…”

Climate policy and induced R&D: How great is the effect?

“…In contrast, Gerlagh (2008) assumes R 1 is equal to approximately 0.63% of world output. 18 Both Popp (2004) and Gerlagh (2008) assume that firms are able to appropriate only 25% of the total benefits of their R&D. 19 In fact, Popp bases his measure of Δ 2 W on discounted consumption, rather than W. Fussel (2007) shows that this can lead to important inconsistencies, and therefore we base our measure on W.…”

Probability distributions for carbon emissions and atmospheric response

“…It is actually the labor fraction of production ω = 1 − α that is estimated from data on labor fraction of compensation (Gollin 2002). What are appropriate methods for derivation and appropriate application of probability distributions for some of these parameters is a question not without controversy, as discussed for example by Füssel (2007). The approach taken here, and described in more detail in Singer et al (2007), is to be as transparent as possible about the data bases and methods used, leaving the reader to decide whether these or alternative approaches are the most useful for any particular application.…”

“…In practice the estimation method used here also raises many questions (c.f. Füssel 2007). Are the answers to questions posed in different languages are really commensurate and representative of what economic decisions attempt to discount and maximize?…”