Nonlinear models and linearly optimal policies: An evaluation

This paper examines the use of linearly optimal policies in systems which are in fact nonlinear. Any linear representation of a non-linear system involves some loss of information about the system, and this loss of information in general leads linearly optimal policies to differ from corresponding nonlinearly optimal policies which are truly optimal. This paper indicates the nature of the resulting discrepancies in a stochastic framework and reports on results of several numerical experiments dealing with macroeconomic stabilization policy.