Practical

Anderson's papers describe a method for solving linear saddle point models. The numerical implementation of the algorithm has proved useful in a wide array of applications including analyzing linear perfect foresight models, and providing initial solutions and asymptotic constraints for nonlinear models. In contract, although the symbolic algebra version has been available for years, few researchers have used it in their work. Most researchers believe that symbolic algebraic solution of useful rational expectations models remains beyond the capacity of present day computers. This papers provides practical advice on how to use symbolic algebra packages to solve moderate sized linear rational expectations models. The computational bottleneck lies in the determination of the invariant space associated with large roots of the transition matrix. This paper provides strategies for avoiding the solution of high order polynomial characteristic, and provides techniques for "switching" to numeric computation at strategic points. With present day PC's, one can solve linear models with ten or twenty equations using these strategies.