Parameter estimation and measures of fit in a global, general equilibrium model

Jing Liu, Purdue University

Abstract

Computable General Equilibrium (CGE) Models have been widely used by researchers around the world for quantitative analysis of economy-wide shocks. However, CGE models are often criticized for resting on weak empirical foundations. Among others, the arbitrary choice of values for key elasticities and the lack of validation procedures in CGE modeling have caused serious concerns for the reliability of the simulation results obtained from models. This is particularly true when model conclusions are sensitive to the values for key elasticities. Data limitations, the numerical complexity arising from extreme nonlinearity in CGE models, and the lack of appropriate measures of fit in the CGE context are all reasons for this dearth of empirical work. This thesis presents an approach to parameter estimation and establishes goodness-of-fit measures for the Global Trade Analysis Project (GTAP) model. It tries to answer two questions. First: what are the most sensible values of trade elasticities given the calibrated structure of the GTAP model? Second: how should we gauge the empirical relevance of this model? An econometric model is linked to the CGE model where a stochastic error is introduced to motivate the goodness-of-fit measures. An approximate likelihood function is employed to measure the size of stochastic errors required to augment the predicted data (multivariate) by the model such that sum of them matches the second moment of the real data. The set of optimum elasticity values is obtained by maximizing the approximate likelihood function in the context of a backcasting exercise. This approach enables us to discriminate among alternative sets of parameter values and among alternative functional forms. In this way, the established measures of fit can be used to improve the infrastructure of the GTAP model. In addition, by drawing on statistical techniques such as Monte Carlo tests, Bootstrapping Resampling, and Permutation experiments, this study is able to judge the robustness of the estimates and to offer the ranges of parameter values for use in global, general equilibrium analysis. The ideas in this study are not limited to the use of the GTAP model. They can be easily adapted for use in the estimation of other multi-region and multi-sector CGE models.

Degree

Ph.D.

Advisors

Arndt, Purdue University.

Subject Area

Agricultural economics|Economics

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS