A robust method for simulating forward-looking models

Methods of computing the type II iterations involved in extended path algorithms for the solution of economic models with forward-looking expectations are described. Particular attention is directed at a method involving the application of a Newton algorithm to a 'stacked' equation system that includes a separate equation for each endogeneous variable at each time point. Analytical results that establish convergence properties for linear models are presented. Both the analytical results and empirical comparisons of the stacked Newton method and the popular Fair-Taylor approach indicate that the stacked Newton method is a viable alternative for medium-sized economic models. (C) 1998 Published by Elsevier Science B.V. All rights reserved.