The simulations conducted for generating IRFs were explosive


I have a nonlinear model and I would like to compare the IRFs under optimal policy versus a simple Taylor Rule.
When I try to run the ramsey policy I received this message:

stoch_simul:: The simulations conducted for generating IRFs to eps_irstar were explosive.

So here are my questions.

  1. Is there any problem with the model? For example: I am not sure that the calculation of steady-state (with an external function) is suitable for Ramsey problem since the steady-state can change under optimal policy.

  2. If it is not, then I read that the solutions are : use a small size of shocks or use a first order simulation. But since my objective is the IRFs, then neither is appropriate for my problem, right? Is there any other solution?

  3. I am using the nonlinear model because when there are distortions like monopolistic competition, the OSR function is not suitable for the model. Is there any other difference between OSR and ramsey policy? I read that the ramsey calculates a condition welfare (starting on steady-state) and OSR unconditional, but this is not completly clear for me.

Thank you in advance.

test.mod (10.4 KB)
root_MBZL44_S04.m (3.2 KB)

  1. Usually, you would enable the pruning option.
  2. For OSR, the planner is restricted to a particular functional form of a policy rule. With Ramsey, the planner chooses optimal policy without being bound by such a restriction.

Thank you for answering, Professor Pfeifer.
May I do some comments on your answers and questions related?

  1. Got it!

  2. This was clear for me. My point wasn’t accurate. My question is: is there a problem using an OSR with a linear model in presence of monopolistic competition and calvo pricing?

  3. I understood that maximization of welfare couldn’t be done with OSR, but I didn’t understand if it is a limitation due to a particular form of a policy rule (as you said) because one must write the problem recursively or is there anything else in addition? Has the distorted steady-state some relation with that ?

  1. The restriction of the OSR-command is that it only allows for purely quadratic objectives, no linear terms. But with a distorted steady state, there would usually be linear terms in the objective function. If you don’t care about those, you can use it.
  2. The implementation of the OSR-command is based on a linear quadratic setup where the goal is finding coefficients that maximize the objective. The first restriction is therefore that you cannot have linear terms in the objective that would require a second approximation. The second restriction is that you are maximizing the objective based on a given rule instead of directly maximizing it.

Thank you, professor Pfeifer. I’ve been thinking of your answer and I think I understand it now.