In modern macroeconomics, empirical models like VAR models can be used as a selection criterion to judge between two or more competing theoretical-based models like DSGE models. Gali (2015) does that in his book. Some of his DSGE models in the book are able to generate the empirical results shown in chapter 1, but only in a qualitative sense (not quantitatively).
Can such DSGE models still be used for policy analysis? I mean, it does a good job qualitatively but does a poor job quantitatively. Such models are still useful, right?