Die asymptotische Verteilung des Prognosefehlers bei Prognosen mit Hilfe eines dynamischen ökonometrischen Modells höherer als erster Ordnung (Q797950)

Based on previous work on the asymptotic distribution of forecast errors in simultaneous equation models - either in their structural or in their reduced forms - the author extends the analysis to the case in which lagged endogenous variables up to order p appear in the model as additional regressors and in which interest concentrates on the l-step ahead forecasts of the endogenous variables. The author derives the asymptotic distributions of the forecast errors, firstly for the m-component vector of forecast errors (when there are m endogenous variables in the model) and for the whole matrix of forecast errors, the joint distribution of the p vectors of forecast errors. A distinction in made between unrestricted and derived, restricted reduced form parameter estimates. In the latter case the reduced form coefficients are derived by standard formulae given a priori knowledge of the structural parameters, whereas in the first case the reduced form coefficients are obtained directly by applying OLS. Within this context questions of asymptotic efficiency of the forecast estimates are addressed. With respect to the efficiency of the forecasting estimation techniques a result is obtained similar to the known results in estimation of simultaneous equations models. Finally, a predictive test is suggested. Proofs are included.