Alternative Approaches to Modeling Time Variation in the Case of the U.S. Real Interest Rate.

This paper compares three approaches for modeling time variation in the U.S. real interest rate: a three-state Markov switching model as estimated by Garcia and Perron (1994), a random-walk model with two-state Markov switching variance, and a time-varying parameter model with two-state Markov switching variance. The findings are generally supportive of modeling continual change in the mean of the real rate process rather than employing a model that limits variation in the mean to a specified number of states. Citation Copyright 1998 by Kluwer Academic Publishers.