Impuse response figure

hi
i have run my model which is a deterministic model and dynare gives me the steady-state but the problem is it doesn’t give me IRFs figure. i appreciate if someone help me
thanks
Ati

In a deterministic model, there are no impulses and therefore no IRFs.

The `simul` or `perfect_foresight_solver` command do not display IRFs, but can obtain the same information by defining shocks in period one (using the `shocks` block). You will need to call the perfect foresight solver for each shock.

Best,
Stéphane.

@stepan-a You will get a simulation with a response to a shock, but I would not call that an IRF. IRFs are usually defined as the change in a variable for a change in the shock, i.e. a partial derivative, given a particular conditioning set. The perfect foresight simulations give you the response to a shocks, but will usually also consider a transition from the initial values to the terminal values, something which is not controlled for. In a sense, you have the same problem as for GIRFs in nonlinear models where you need to subtract the baseline without shock to get a proper IRF. Only if initial and terminal values are the steady state, the two coincide (assuming a conditioning set of 0 for all other shocks at all times).

1 Like

@jpfeifer If the initial condition for the states is the steady state, there is no difference with IRFs (assuming the steady state does not move for other reasons than the shock on the innovation). In this sense, we can call this an impulse response function. We perfectly control for the transition from the initial condition, since we are able to choose the initial condition (with the `initval` or `histval` blocks). Choosing the steady state as an initial condition, we kill the transition effect. Also, as you say, we can consider a baseline scenario without the shock (only with the transition effect) and compute the difference with a scenario where a shock is added. Depending on the initial state, we will not observe the same marginal effect of the shock since the model is nonlinear. We could, I suppose that it would be closer to GIRFs, consider a distribution of states and report the average transition to the steady state. But obviously the result would depend on the choice for the distribution of the initial condition (even if consider deviations to a baseline scenario).