Essays on Climate Change Computable General Equilibrium Models

Abstract

This research reviews the problems of conventional computable general equilibrium (CGE) models which are widely used for climate change policy analysis. To solve the problems, it proposes multivariate distribution approach as an alternative way of representing the production activities in model structures and assesses the possibility of its practical employment.

In the first part of this research, the basic characteristics of three well known global CGE models are reviewed and production function structures are pointed out as the main sources of the differences in carbon emission projections among models. Two experiments are introduced regarding the effects of changes in production function structures. In one experiment, the nested structure of constant elasticity substitution (CES) functions is substituting with alternative nesting structures. In another experiment, fixed input structures are partly applied for incorporating bottom-up approach with top-down mechanism of CGE models. The results show that these structural changes cause a considerable impact on the prediction results of greenhouse gas emissions and carbon prices. Also, the experiments are extended to the comparison of GDP losses among different model structures. Simulations for the case of Korea reveal that the estimations of GDP loss differs among model structures, raising some issues on applying them into practical policy making.

In the second part, the performance of a global CGE model is analyzed in marginal abatement cost estimation when data disaggregation is applied. Extraordinary carbon prices are reported for the case of relatively large share of capital in the economies of a few developing countries. Empirical evidence indicates that the abnormal phenomenon is accounted for by the proportional relationship between capital intensity and carbon price. The analysis is extended to CES functions with a numerical analysis, concluding that the unusual phenomena may be connected to distribution parameters of CES functional forms which are most widely used in CGE models.

In the last part, multivariate distribution approach is applied for an alternative description of energy related production activities. Applying theories on the microfoundations of aggregate production functions, it is shown that a set of bottom-up microscopic information can converge to specific aggregate production functions if assumptions are imposed on the statistical distribution of local production technologies. The actual characteristics of statistical distributions were reviewed for a real dataset of energy intensive manufacturing sector of Korea. To facilitate simulations and conveniently reproduce the relationships embedded in multivariate joint distribution maps, a statistical tool called copulas is introduced in advance. After the basic theory of copulas is briefly introduced, the performance of a copula model is investigated, revealing that a copula model is successful in describing heterogeneous microscopic information. After the introduction of copulas, a new type of CGE model is applied, in which an aggregation of local Leontief production functions takes over the role of conventional global production functions. A pilot model is composed to apply this scheme to a CGE model and it is shown that this new approach has some advantages: it eliminates the effect of the past time data and improves the precision of projection results.