Estimating DSGE Models with Long Memory Dynamics

Recent literature claims that key variables such as aggregate productivity and inflation display long memory dynamics. We study the implications of this high degree of persistence on the estimation of Dynamic Stochastic General Equilibrium (DSGE) models. We show that long memory data produce substantial bias in the deep parameter estimates when a standard Kalman Filter-MLE procedure is used. We propose a modification of the Kalman Filter to effectively deal with this problem. The augmented Kalman Filter can consistently estimate the model parameters as well as produce more accurate out-of-sample forecasts compared to the standard Kalman filter