Bayesian Local Likelihood Estimation of Time-Varying DSGE Models: Allowing for Indeterminacy

This paper modifies and employs a Bayesian Local Likelihood approach to estimate time-varying parameters of a New Keynesian model and assess such time variations using US data. Our modification contributes to the expanding literature by novelly integrating indeterminacy into the time-varying estimator. Further, we implement a one-step approach based on a unified solution set obtained by the augmented linearized rational expectation model simultaneously allowing for both determinacy and indeterminacy. The evidences suggest substantial time-variations in many parameters, particularly those associated with the Fed monetary policy rule and characterized by volatilities in the economy. This study also shows that allowing time-varying parameters improves density and point forecasts in comparison to a fixed-parameter DSGE model. We investigate implications of the time-variation for monetary policy effectiveness and find that the increase in the policy response to inflation from the pre-1979 to the post-1982 alone does not suffice for explaining the U.S. economy’s shift to determinacy, unless it is accompanied by either the estimated decline in trend inflation or the estimated change in policy responses to the output growth.