shocks

[martom]

Dear Michel,

Thanks so much for your last very helpful and super prompt reply.

Another quick question.

From what we understand, specifying a shock with a var-cov matrix will shock the system each period with a shock drawn from a normal distribution with parameters of the matrix.

We would instead like to run impulse response functions of the following nature:

1) shock the system with a temporary shock (ie, a one period shock than nothing thereafter)

2) shock the system with a permanent shock (ie, a one period shock that retains that value thereafter)

We've looked in details in your examples and doc, but find nothing pertaining to these questions.

Thanks again for your help.

[MichelJuillard]

Dear Michel,

From what we understand, specifying a shock with a var-cov matrix will shock the system each period with a shock drawn from a normal distribution with parameters of the matrix.

We would instead like to run impulse response functions of the following nature:

1) shock the system with a temporary shock (ie, a one period shock than nothing thereafter)

2) shock the system with a permanent shock (ie, a one period shock that retains that value thereafter)

We've looked in details in your examples and doc, but find nothing pertaining to these questions.

Thanks again for your help.

A key question is what do you assume concerning agents expectations:
do they know that there will be nothing thereafter or that the shock will be permanent?
If you assume that they do know, then you model is deterministic and you should use 'simul' and 'shocks' to specify the temporary shock. Use 'initval' and 'endval' to specify the permanent shock.

If the agents believe that the system is stochastic but you want to study the consequences of one surprise shock, look at the IRFs provided by stoch_simul. This is the response to one one-standard-deviation shock in first period.

If the agents know that this surprise shock will be permanent then you have a nonstationary model. You should stationarize it first (look at the topics on nonstationary models)

Best

Michel