Computing estimates of continuous time macroeconometric models on the basis of discrete data

Computational aspects of obtaining estimates of continuous time macroeconometric models on the basis of discrete data are examined. In contemporary, dynamic disequilibrium models, the central feature is shown to involve reliably computing the exponential of certain block triangular matrices. Owing to their non-normality, there is no universally robust method of doing so. Four methods of computing the matrix exponential are compared and contrasted in context, in terms of reliability, accuracy, efficiency and stability. These are then applied to examine the robustness of computing estimates of a prototypical continuous time model based on maximising a Gaussian likelihood. (c) 2004 Elsevier B.V. All rights reserved.