Stochastic Extended Path Question

Extended Path
1

Hello,

I wanted to know if it’s possible to conduct SEP while going from one SS to another. I initially solved with the perfect foresight solver for a permanent shock that goes from one SS to another. In that transition, is it possible to solve under SEP (random shocks + agents being aware of it).

Thank you,

2

My understanding is that should be possible. Do you experience issues?

3

Hi, I am sure it’s possible but I believe we do not have an interface for such a setup. Basically, assuming that agents learn in period 1 that the economy goes to a new steady state, you need to

  1. Compute the first steady state, and store it in a vector (say yinit)
  2. Change the values of some parameters
  3. Compute the terminal steady state (with the steady command)
  4. Manually set the options for SEP in options_.ep
  5. Use the extended_path MATLAB routine (not the Dynare command) and pass yinit.

Obviously steps 4 and 5 require knowledge of the Dynare internals… But it’s not hard to figure out how to do it.

Depending on the distance between the two steady states the algorithm may fail in period 1. Because the nonlinear solver will use the terminal condition in all periods as an initial guess. It must be possible to tweak the code and use the solution of the perfect foresight simulation of the transition between the two steady state as an initial guess.

Best, Stéphane

4

Thanks a lot Stephane and Johannes. I will try that and let you know.

Best