Uncertainty of policy recommendations for nonlinear econometric models: some empirical results

A method for evaluating the reliability of policy recommendations derived from a linear dynamic structural econometric model in the framework of the linear quadratic control problem has been recently proposed by Friedmann (1980, 1981). The method analytically derives the asymptotic distribution of the estimated optimal policy and in particular the asymptotic standard errors of policy instruments, with respect to structural coefficients estimation errors. The use of analytic simulation and of Monte Carlo techniques allows to extend Friedmann's findings to medium and large size dynamic linear models and to nonlinear econometric models. Empirical results for some nonlinear models of national economies are reported in the paper.