# Deterministic model - perfect foresight simulation

Dear All
I have a relatively standard DSGE model which I solve under commitment, with all the standard commands - no problems. One of shocks there is a technology shock with z = rho*z(-1) + e;

Now, I want to produce a deterministic simulation, namely start with z(-1) = 0.01, and compute z = rho*z(-1) and paths of other variables, i.e. no e=innovation.

What should I change? I read help on perfect-foresight, but it lacks examples. The example which I found - is not liked by dynare.

My understanding is the following:
I keep
shocks block, but there

shocks;
var z;
periods 1:1;
values 0.01;
end;

Here dynare does not like values, I think, but what do I need to say to generate the initial deviation of z?

Now, I keep ramsey_model command in my file.

then what?
perfect foresight_setup(periods=50):
perfect_foresight_solver(lmmcp);

I did that - it produced me something even without me giving the shocks block! Is it trying to solve finite horizon problem? This is way too much for me - I have a DSGE model so I need the “usual” steady state, and simply a deterministic path towards this steady state. I am more than happy that the model would be log-linearised first (in fact this is what I want). Is there any way to get this type of a solution?

If this is impossible, where transition matrices are stored if the stoch_simul command is used? So I can generate such path and plot it myself? I am aware of oo_.dr.ghx and it would be fine to work with it, if not Ramsey that I need. The columns and rows are all mixed up there. If what goes on the screen had a solution for LMs - this would be a great help, but it does not! On the screen I have all states listed, but not solutions for all states (for LMs), and it is very difficult to extract solutions for LMs from oo_.dr.ghx even if you provide some columns with permutations - I am getting old and find it difficult to reconstruct.

I suspect I am doing something totally wrong and waste hours and days to reconstruct either the transition matrix or just simply plot a deterministic paths. Please help!

Why do you want to do a fully nonlinear deterministic simulation in the first place? Why don’t you simply use the `simult_`-function to do the simulation you desire?
See e.g. RBC_state_dependent_GIRF.mod that generates an IRF at capital 10% below steady state.

A long overdue thank you! I made both work and the perfect foresight and simult_ where needed. Pointing out this command was very helpful. Many thanks again!