Forecasting with the input-output model: some econometric problems

Abstract

The starting point of this paper is a review of the standard input-output forecasting methodology and of some of the developments in the stochastic context. This review shows the basic hypothesis and the limits of the traditional methodology. In order to avoid these limits the problem is reformulated in the context of statistical inference, introducing the concept of estimate of the matrix of technical coefficients and deriving the variance-covariance matrix of the reduced form coefficients (Leontief-inverse) and of the prediction. In this context the traditional components of the prediction error appear as nothing else than neglected estimation problems.