Estimating Parameters and Structural Change in CGE Models Using a Bayesian Cross-Entropy Estimation Approach

This paper uses a two-step Bayesian cross-entropy estimation approach in an environment of noisy and scarce data to estimate behavioral parameters for a computable general equilibrium (CGE) model and to measure how labor augmenting productivity and/or other parameters in the model shift over time to generate historically observed changes in economic structure. In this approach, the parameters in a CGE model are treated as fixed but unknown, which we represent as prior mean values with prior error mass functions. Estimation of (and inference about) the parameters involves using an information-theoretic Bayesian approach to exploit additional information in the form of new data from a series of Social Accounting Matrices (SAMs), which we assume were measured with error. The estimation procedure is “efficient” in the sense that it uses all available information and makes no assumptions about unavailable information. As illustrations, we apply the methodology to estimate the parameters of a CGE model using data for South Korea and for Sub-Saharan Africa.