Model Simulation on Parallel Computers

Abstract

The simulation of large-scale macroeconometric models may be considered as a solved problem, given the performance of the computers available at present. This is probably true if we run single simulations on a model. However, if it comes to solving a model repeatedly a large number of times, as is the case for stochastic simulation, optimal control or the evaluation of forecast errors, the time necessary to execute this task on a computer can become excessively large. Another situation in which we need even more efficient solution procedures arises when we want to explore the behavior of a model with respect to continuous changes in the parameters or exogenous variables.