The effects of additional non-stationary processes on the properties of DSGE-models

Abstract

DSGE models are based on the trend-cycle decomposition. The standard approach implies an out-of-model decomposition of the data, in which the trend component is discarded, and the parameters of the model are estimated on the cyclic one. This approach can lead to the loss of statistical information and reduce the quality of the model, which is crucial for practical purposes. The study suggests adding several sector-specifi c exogenous non-stationary processes to the model, which complement the standard DSGE model. The in-model detrending is described, and an approach to GMM-estimation of the non-stationary processes’ parameters is proposed. Several results are obtained. First, the inclusion of such non-stationary processes in the model increases the marginal density and improves the accuracy of forecasting within the sample. This result is robust to the inclusion of measurement errors in the model. Secondly, it is shown that the addition of exogenous trends allows obtaining a more plausible decomposition of data into a trend and a cycle. Finally, the use of the GMM approach to estimating the trends’ parameters makes possible to increase the marginal density. The results obtained in the paper can be used to create practice-oriented DSGE models.