Model evaluation

Hello, I am trying to quantitatively assess my simulated DSGE model in fitting key macroeconomic statistics. I have a few questions:

  1. Is it correct to compare the standard deviation of the cyclical component of key macro variables in the data with the standard deviation that I see under “Moments of simulated variables” after running the code, to say something about the fitting properties of the model?

  2. How do you assess whether these standard deviations are “close enough”?

  3. Do you have any clue of why the volatility of investment that I osberve under “Moments of simulated variables” is very small, whreas it is generally the most volatilile component of GDP? Which parameters, generally speaking, might affect this in a standard DSGE model?

Thank you!

  1. In principle, yes. But you need to make sure that your are comparting apples to apples, not oranges. In particular, apply the same filter to your model variables as to the data (e.g. first differences or HP-filter)
  2. There is no guideline. A cynic would say: anything the referee accepts is fine. Without doing proper inference, there is no way to judge the model fit.
  3. That may have to do with different filtering (or a wrong FOC in the model). Regarding parameters: the investment adjustment costs, obviously. The TFP persistence may also have big effects.