I have a non-linear DSGE model that IRFs results are compatible wiht theory. for Assessing of DSGE results robustness , I use “log” of real data and deterend them whith Hp filter. then, I calulate mean and variance of cycle series and compar with that of Dynear report. but there is a significant diference between them! it is nessecry to say that may time series data are short (20-30years). my question is:
is hp filter mandatory for dsge model to assess robostness? if not, is there any article for reference to it?
thanks in advance.
No, HP filtering is not necessary. But often it’s a convenient way to summarize the cyclical properties of the model. In any case, it’s important to treat the data and the model results equally. Have you also HP-filtered the logged variables in the model for your comparison?
Once logged variables was hp filtered and secondly level variables was hp filtered. Both of them was not similar to DSGE results.
Then there is most likely still a mistake in your model. One issue may be the shock variances.