A comparative study of the empirical impact of the joint estimation approach and the alternative selection approach for application to econometric models with a pure autocorrelation structure : the case of the General Fechner-Thurstone direct utility function

Abstract

This dissertation is an empirical, comparative study of the impact of two different approaches, the joint estimation approach and the alternative selection approach, in dealing with econometric models of a typical autocorrelation structure in the context of the consumer variable preference theory under the General Fechner-Thurstone direct utility function. For both the first- and second-order processes, several widely-applied joint estimation methods have been chosen for study and comparison with the alternative selection method on the basis of two empirical models. The different approaches are evaluated from both a statistical and economic viewpoint for five commodity groups. It is found that the specified maintained hypotheses from the two approaches can be significantly different. Further, it is concluded that the alternative selection approach is more favorable than the usual joint estimation approach in terms of the Durbin-Watson statistic, the adjusted multiple correlation coefficient, and the condition index. Noticeably different results are observed in the test of homogeneity properties of marginal rates of substitution and elasticity estimates based on the two approaches.