Evaluating welfare losses due to the use of approximated stochastic optimal control algorithms: An application to the banca d'italia quarterly model

Stochastic optimal control of large-size nonlinear econometric models can require enormous computing resources. As an alternative to the full stochastic approach, the standard deterministic control (neglecting the stochastic nature of estimated models) is often used. In recent articles, Hall and Stephenson have proposed an algorithm that takes into account, at least partially, the stochastic nature of optimal control with econometric models, without calling for enormous increases in computing time. Their algorithm, however, neglects the role of the variance. In this paper, the full stochastic optimal control algorithm described in detail in Cividini-Siviero (1992) is used to solve realistic problems of stabilizing the Italian economy towards a steady-state growth path. All problems use the version of the Banca d'Italia Quarterly Model that has been modified so as to ensure the existence of a long-run equilibrium. The welfare losses due to the approximated algorithms are then compared for two different optimal control exercises.