Does trend inflation make a difference?

Although the average inflation rate of developed countries in the postwar period has been greater than zero, much of the extensive literature on monetary policy has employed models that assume zero steady-state inflation. In comparing four estimated medium–scale NK DSGE models with real and nominal frictions, we seek to shed light on the quantitative implications of omitting trend inflation, that is, positive steady–state inflation. We compare certain population characteristics and the IRFs for the four models by applying two loss functions based on a point distance criterion and on a distribution distance criterion, respectively. Finally, we compare the RMSE forecasts and we consider also an indirect inference test. We repeat the analysis for three sub-periods: the Great Inflation, the Great Moderation and the union of the two periods. We do not find strong evidence that a model with trend inflation should always be preferred. During periods of high inflation or when a backward-looking component, indexed to past inflation, is not incorporated in the model, using a model that employs trend inflation can improve the analysis. Nevertheless, where there is uncertainty concerning the change of an inflation regime, such as the recent drop, we suggest adopting a traditional approach that does not use trend inflation.