I’m trying to replicate a paper but with data for a different country. I tried to simulate the model first, with the parameter estimates given in the paper, to check whether the model was correctly written. Then i try to estimate it… and i get that with these initial values (the same as in the simulation) it’s not possible to compute the likelihood. Is that possible, or did I do some mistake in writing my code? I checked many times, but i can’t find the problem.

No, that should not be possible. Did you really provide explicity starting values for estimation or do you use Dynare’s default values, i.e. the mean of the prior distribution? If in doubt, post the mod-file.