In a recent note [Rouzier (1973, a)] we indicated a method for correcting certain parameters of an econometric model in order to minimize a weighted sum of mean square errors for the endogenous variables. A crude justification of this method was that, since the equations are generally estimated separately, there is no obvious reason why the complete model should lead to consistent solutions. Indeed, when a model is being constructed, there are implicit interactions whose effects are not revealed until numerical simulations are performed.
The final run of a model differs a lot from the first one because the builder adds subjective informations to the results of the estimation. But when changing a coefficient, or when opting for another specification, he is probably paying a price in terms of the economic meaning of certain trade-offs. This is a rather vague statement, but we think that this study is able to contribute to a more complete understanding of the mechanisms of our models.
In this paper, we intend to show that the above mentioned method can in fact be used not only to improve the performances, but also as a tool to analyze the trade-offs of a model.