Issues with the IRFs

Stochastic simulations
1

Hi,
I am trying to understand how to get the IR right for the variable POstar (international price of oil).
In the first model (… denoted Model5_externaldraft3), the IR does not show up. I though it has to do with the overall external sector part of the model. So I expanded with the all relevant variables in the new model (Model5_external). Now the impulse is not smooth.

My primary focus is Model5_externaldraft3 as it is more simplified. I will appreciate any advise to get the impulse showing. Thanks.

Model5_Externaldraft3.m (4.1 KB)
Model5_Externaldraft3.mod (7.6 KB)
Model5_External.mod (8.7 KB)
Model5_ExternalSS.m (4.2 KB)

2

I don’t understand what the problem is. That most variables do not react to epsPOstar? You can check that E move exactly opposite to POstar to leave PO (and the rest of the model) unaffected.

3

Thanks @jpfeifer . I am trying to get the IRF showing on the simple version of the model (…draft3). You can see that when running there no IRF displayed.

4

This version shows some IRFs. Note that most IRFs are 0 and therefore not shown.
Model5_Externaldraft3.mod (7.6 KB)
You could use irf_plot_threshold = 0 to change that.

5

Thanks a lot @jpfeifer. Is there any way to make the IRF more smooth? Or is it normal? I find it awkward.I tried changing the shock parameters but it does not help much.

6

Have a look at the size of the response. If you set

there will often be wiggles due to numerical error as the IRFs are not exactly 0 but so small any wiggles are meaningless.

7

Thanks @jpfeifer. I have been looking for any published paper with these wiggles but I did not find any for now. Not sure if moving on with the model expansion will be fine, in terms of publication. I again reduced the model and the equations and noticed that the wiggles comes from the introduction of oil price shock : See lines 172 and 235. Not sure where could the error arise. Do you mind scanning the attached ? Thanks.

Model5_Externaldraft3.mod (7.7 KB)
Model5_Externaldraft3.m (4.1 KB)

8

Those IRFs are simply 0. They only show small movements due to numerical precision. So your problem is one of the shock not affecting the relevant variables.

9

@jpfeifer Thanks. I simplified the model to a working level to align with the source paper, and noticed that oil price shock was the main cause. I removed it. I have been trying to reincorporate it back but the same issue arises. I either get zero /nearly zero impulses, or # equation greater than # variables (removing one equation does not help as the BK rank condition fail to verify). Do you think this has to do with the way shock to oil price shock should enter the model or it could be the whole model itself or the parameters. I initially introduce oil shock in this way. But this does not help: Thanks
Model5_Externaldraft3.m (4.1 KB)
mod5_no_oil_shock.mod (7.3 KB)

log(PO)=(1-rhoPO)log(POss)+rhoPOlog(PO(-1))+epsPO;

10

Oil prices are an exogenous shock here. So the issue is the transmission into the model. You think harder about the actual economics of what is going on in your model.

11

Thanks so much @jpfeifer for your great support.