“Suppose U is my utility function, if I want to compute conditional expectation of utility, should I just create a new variable con_util = U + bet* U(+1), where bet is the subjective discount factor, and then just read the mean of con_util from the results reported by dynare in a second order approximation?”

You answered yes to this question. I am wondering if this measure of conditional expected utility can be used to rank the welfare under different monetary policies?

I was wondering if you have solved your problem in the meantime because I was posing myself the same question about welfare ranking.

Would it not be more appropriate to look at the IRFs of the “con_util” variable when comparing different parameter configurations and see which one has the highest amplitude (lowest negative amplitude)?

I have a situation where the ranking of the theoretical means implies something else in terms of utility losses to me than the IRFs show me at a first glance. It is difficult to see because the graphs cross and therefore I was wondering if it is possible to calculate the integrals of the so constructed con_util IRF.

If anyone has any suggestions about welfare comparison (pure utility comparison without a loss function or so) I would be very thankful about comments.

I am wondering whether we can simply compare the con_util at period one. Since con_util = U + bet* U(+1), so con_util(1)=U(1)+betU(2)+bet^2U(3). If we want to compare the welfare conditional on that the initial state is at the steady state, then con_util at period 1 just measures that. How do you think?