Title: Non-stationary DSGE models with time-varying steady state
Authors: Viktors Ajevskis - Bank of Latvia (Latvia) [presenting]
Abstract: DSGE models are designed to explain cyclical features of the data. There are two approaches to deal with data for estimating the parameters of DSGE models: a. filter the data using statistical filter and then estimate the structural parameters with the output of the filter; b. transform the data using model-based specification of what the non-cyclical component is, then estimate the structural parameters with the transformed data. In both cases it is assumed that there exists a fixed steady state (either in the level or in the growth rate) in the model and the Blanchard-Kahn conditions hold. However, both approaches have their own problems. In the proposed approach a DSGE model with a unit root process in technology is considered. In this case the fixed steady state solution does not exist. A time-varying steady state is defined as a solution to which the economy converges in the absence of the future shocks. Presenting the data as a sum of the time-varying steady state and deviation from that allows using the Kalman filter for estimating the model parameters and unobservable variables. A prototypical DSGE model and the US data to estimate the model parameters and the output gap is used.