Finite-Sample Properties in Stochastic Predictors in Nonlinear Systems : Some Initial Results.

Many econometric models for forecasting and policy analysis consist of a statistically estimated system of nonlinear stochastic equations. The distinguishing feature of these models is the nonlinearity of the solution for the endogenous variables in terms of model disturbances. Despite the widespread use of these models, there has been little formal analysis of predictions based on such models. Furthermore, practitioners' validation of such models has proceeded, for the most part, on a informal basis.