Solving DSGE Models - When Local Approximations Fail
57 Pages Posted: 26 Jun 2017 Last revised: 18 Oct 2018
Date Written: December 01, 2017
This paper studies the effect of persistent growth risks on the solution accuracy of dynamic stochastic general equilibrium models. We find that whenever the economy is exposed to risks with long-lasting effects a local Taylor-expansion based solution method does not suffice. In particular, this affects asset pricing quantities and welfare implications. We compare perturbed solutions for various economies with solutions obtained with a projection algorithm. It is shown that besides slightly misstating macroeconomic moments a perturbed solution strongly understates the mean risk-free rate and the wealth consumption ratio. Further, we identify parameters driving the approximation error and compare different degrees of approximation. Finally, we show that a Chebyshev projection does a better job at approximating asset pricing and welfare quantities than a Taylor-expansion even for low order polynomials.
Keywords: Growth risks, DSGE models, solution methods
JEL Classification: C63, D58, G12
Suggested Citation: Suggested Citation