I have a simulated model and now I am trying to derive the optimal policy rule. I ran the optimal simple rule in dynare as suggested in the guide but it reproduced the initial values of the parameters as the optimal parameter which is absurd.
OPTIMAL SIMPLE RULE
OSR: Initial value of the objective function: 2.40758
f at the beginning of new iteration, 2.4075798109
Norm of dx 0
Improvement on iteration 1 = 0.000000000
improvement < crit termination
OPTIMAL VALUE OF THE PARAMETERS:
Objective function : 2.40758
Can anyone please suggest what is the issue?
utility_opt_10.mod (9.26 KB)
You need to check your model. I tried how different values for the parameters affect the moments of the objective variables using
stoch_simul(order=1,irf=0,nocorr) pie r_ib;
and there is no movement. Given that the variance of these variables does not move when you change phi_p, it is hardly unusual when the optimizer returns the starting values as the parameter seems to be not identified.
I have a follow up question. I am actually getting the same problem. The optimized parameters values are the same as the original values.
Prof. jpfeifer could you please tell me what you mean by “check your model” what should one check exactly?
More specifically, for my case I am looking for optimized parameters that can maximize welfare.
Using different values of the parameters gives me slightly different variances of the variables.
Thank you for your help.
Without the mod-file, this is hard to say. Have you tried the unstable version with a different optimizer?
Thank you for your quick response.
I just tried the unstable version, I get the same thing. .
I searched more on the forum. I found few more examples.
I think I have a wrong the optim_weight block.
My understanding is that if for example I have
r = (yh/yh_s)^Gamma1*(qh/qh_s)^Gamma2;
the optim_weight is :
osr_params Gamma1 Gamma2;
Where is my mistake at that point? do I need any other variable?
unfortunately, I can’t post the file, it is a join work. I am not allowed.
Please help me with a description on how to write that part I think it will change something.
Thank you so much again.
Then simply send me the files via email or PM.