Gains From Employing Sparse Matrix Techniques in the Anderson-Moore Algorithm

The Anderson-Moore algorithm is a powerful method for solving linear saddle-point models. The algorithm has proved useful in a wide variety of applications, including analyzing linear perfect-foresight models and providing initial solutions and asymptotic constraints for nonlinear models. The algorithm solves linear problems with dozens of lags and leads and hundreds of equations in seconds. The technique works well both for symbolic and numerical computation. The existing implementation of the algorithm exploits aspects of the inherent sparsity of the linear systems that alternative approaches cannot. However, incorporating sparse matrix storage and linear algebraic routines dramatically improves the scalability of the existing implementation. This paper describes the new implementation and documents the improvements in performance. The paper presents numerical results for solving a large macroeconomic model. The author can provide potential users with a C version on request.