Solving DSGE Models Without a Grid

Abstract

This paper presents a global solution method to DSGE models, which does not depend on a grid and hence does not suffer from the curse of dimensionality. The method enables to approximate the Taylor series of the policy function at any arbitrary point of the state space. Once the Taylor series is approximated at a given point, the constant term of the series provides the model solution at that point. Since the solution is not based on a grid, the computational costs are significantly lower compared to grid-based methods, because the model is solved only at points of interests (e.g. along a simulation path). Accuracy is high, compared to other methods, and it improves significantly by discretizing time into short periods.