A BAYESIAN ALTERNATIVE TO GENERALIZED CROSS ENTROPY SOLUTIONS FOR UNDERDETERMINED ECONOMETRIC MODELS

This paper presents a Bayesian alternative to Generalized Maximum Entropy (GME) and Generalized Cross Entropy (GCE) methods for deriving solutions to econometric models represented by underdetermined systems of equations. For certain types of econometric model specifications, the Bayesian approach provides fully equivalent results to GME-GCE techniques. However, in its general form, the proposed Bayesian methodology allows a more direct and straightforwardly interpretable formulation of available prior information and can reduce significantly the computational effort involved in finding solutions. The technique can be adapted to provide solutions in situations characterized by either informative or uninformative prior information.