Policy determination through simulation of non-linear econometric models

Simulation of stochastic non-linear econometric models is known to have desirable analytic content only when the error terms affecting the structural equations are incorporated within the simulation procedure. This paper demonstrates that stochastic simulation, repeated solution of a model with error terms explicitly incorporated in quantified form, results in an empirical distribution function for the endogenous variables which converges uniformly to the true distribution function. This result allows the construction of confidence intervals on the paths of the endogenous variables of the model. Furthermore the Bayes' Principle is extended to cover optimal policy determination for a finite set of available policies for stochastic non-linear econometric models; resulting in a practical procedure available for development planning.