How large and uncertain are costs of 2030 emission reduction target for the European countries? Sensitivity analysis in a global CGE model

CGE models are popular tools for assessing the impact of economic (climate and energy, specifically) policy on the economy, yet the simulation results are sensitive to parameters assumed by the modeler. However, econometric evidence on those parameters available in the literature is often scarce or ambiguous, as well as there is difficulty in finding results tailored to a specific CGE model (with its specific sectoral and regional disaggregation, nesting structure production functions etc). In practice this makes a choice of parameter values more or less arbitrary, and in fact in many cases the modelers simply follow the perhaps equally arbitrary choices made by other authors. Although such an approach does not imply that simulation results are meaningless, it calls for at least a clear communication of uncertainties to the reader. In the paper we present mean results of simulation of the 2030 emission reduction target for the EU countries together with standard deviations of mean results. We also discuss sources of parameters uncertainty. In order to investigate model parameters uncertainty we conduct systematic sensitivity analysis based on Stroud's (1957) Gaussian quadratures in a static global CGE model. By “systematic” we mean that alternative values of parameters are picked in a systematic way, i.e. they are determined by means of some specific method in order to explore the whole domain of plausible values. Our preliminary findings can be summarized as follows. First, uncertainty of model simulation results driven by the uncertainty in assumed elasticities is quite remarkable with double-digit variation coefficients in many cases. Second, the uncertainty is larger with respect to non-energy parameters than to energy parameters. Finally, there is a clear pattern with mostly the New Member States experiencing relatively high cost of emissions reduction in terms of GDP and consumption loss. In the extreme case of strictly rigid energy mixes (no substitution at an industry level), these costs roughly double.