Tests and uses of macro-econometric models: A critical survey

Summary In spite of some recent developments, the general area of methods for testing and using macro-econometric models is in a very underdeveloped state. Statistical measures like Theil's inequality coefficient or simple mean square error give some idea about the performance of forecasts of individual variables. They do not give a satisfactory measure for the model as a whole. Even more important, these statistical measures do not provide a proper economic measure of the efficiency of the model. In general we would like to have some measure of the sensitivity of optimum policy developed on the basis of the model. The acceptability of the model will then depend on whether the degree of uncertainty thus associated with the objective function of the policy maker is acceptable. This of course requires specification of the objective function of the authorities using the macro-econometric models. Here again the progress achieved so far is rather limited. We have argued for a programming approach to the problem where instead of searching for a general welfare function, we accept one target as the object of maximisation while other targets appear as constraints. This approach however, is only a shortrun solution. The longer run problem of specifying a social welfare function and evaluating and using macro-econometric models has so far been barely scratched on surface. The author is grateful to Professor J. A. Sawyer for his valuable comments at various stages of this paper. This survey was done in connection with the project on Econometric Model of the U.K. that is going on at London Business School under the supervision of Professor R. J. Ball.