INFORMATION BASED MODEL AVERAGING AND INTERNAL METANALYSIS IN SEEMINGLY UNRELATED REGRESSIONS WITH AN APPLICATION TO A DEMAND SYSTEM

This paper presents an information based model averaging and internal meta-analysis procedure that is easily applied to a large model space. In the application, the procedure is used to investigate the efficacy of some recently contested commodity promotion programs. The investigated model space consists of 576 demand systems. The internal meta-analysis indicates that theoretical restrictions and evaluation points are more important than alternative functional forms and explanatory variables in determining the elasticity values. The model averaging weights strongly support the theoretically consistent classical demand systems without promotion. The weighted or meta-price and meta-expenditure elasticities are presented and discussed.