A remark on econometric modelling, optimization and decision making

Abstract

This paper is concerned with an operational issue about a common type of econometric model viewed as a decision making tool. It is generally recognized that the predictive performance of many presently available econometric models may be considered somewhat below a desirable standard for decision making; indeed facts such as large variances of parameter estimates, fast changes in the system being modelled, presence of random disturbances, and flattening of nonlinearities into linear approximants may be responsible for reducing the usefulness of the model as an opera tional tool in decision making. This paper reports the initial part of an investigation on whether it is possible to improve the predictive per formance of a given econometric model without modifying the form of the model equations. As a first step towards such an end, we introduce here a calibration problem — that is a certain mathematical program —, apply it to a specific econometric model, and report the preliminary results obtained.

This research has been supported in part by the Ministero della Pubblica Istruzione (Art. 286 T.U.).