Meta Response Surface Design for General and Partial Equilibrium Models

Due to the complexity of general and partial equilibrium models, conventional sensitivity analysis, qualitative reviews or literature-based meta-analyses do not allow for detailed assessments of the role of individual parameters and policy shocks across different models. Therefore, the partial equilibrium model “GSIM” and a single country CGE are employed to generate synthetic scenarios based on randomly specified combinations of base data, elasticities and tariff changes selected from previously specified, plausible ranges. The synthetic meta-data has the advantage that the values of explanatory variables are measured exactly. This makes it possible to explore complex issues of functional form and interaction between variables in the estimation of the response surface of each model as well as for a joint response surface of both models. The results indicate that firstand second-order polynomials provide sufficient approximations of the model responses, and especially for the CGE model, interaction terms of elasticities with policy variables play an important role. Furthermore, simultaneous estimation of a response surface of scenarios from both models proves to be feasible and enables quantitative comparisons of different model output, e.g. welfare measures.