VAR Impulse Response Function and exogenous variables

For technical questions regarding estimation of single equations, systems, VARs, Factor analysis and State Space Models in EViews. General econometric questions and advice should go in the Econometric Discussions forum.

Moderators: EViews Gareth, EViews Moderator

karlaborja
Posts: 1
Joined: Tue Sep 13, 2011 4:42 pm

VAR Impulse Response Function and exogenous variables

Postby karlaborja » Thu Sep 15, 2011 1:10 pm

I am using a VAR model with exogenous variables as the main explanatory variables (e.g. the effect of natural disasters on GDP). The Natural disaster variable is a dummy variable with value of 1 every time the region under study experienced a natural disaster.

The objective is to calculate impulse response functions, but such functions are calculated by Eviews ONLY if the natural disaster variable is added in the VAR model as an endogenous variable. There should be a way to impose a restriction on the matrix of parameters indicating the exogeneity of the natural disaster variable but still be part of the VAR system.

How could I run a VAR model and then get impulse response functions from an exogenous variable such as natural disasters?

EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 13305
Joined: Tue Sep 16, 2008 5:38 pm

Re: VAR Impulse Response Function and exogenous variables

Postby EViews Gareth » Thu Sep 15, 2011 1:24 pm

There's nothing built in that will do it for you. However, it seems to me all you really want to do is forecast from the VAR with your dummy equal to zero, then forecast from the VAR with your dummy equal to 1, and then see how they differ?
Follow us on Twitter @IHSEViews

Rob H-D
Posts: 1
Joined: Wed May 19, 2021 3:11 pm

Re: VAR Impulse Response Function and exogenous variables

Postby Rob H-D » Wed Aug 18, 2021 3:35 pm

Hi Gareth,

Is this still the case that there is no function for this in eviews (i saw the original post was 2011)?

I am trying to create historical decompositions and an impulse response function for the endogenous and exogenous variables. I am currently using the eviews 11 university edition.

My SVAR model looking at the crude oil market. This based on Kilian and Murphy's original model (Crude oil price, oil production, global economic activity and crude oil inventory levels) and a later extension (Zhou, 2019) to include the impact of the real exchange rate. However, as with monetary policy variables in most SVAR models it is treated as an exogenous variable.

The suggestion was to run two models one including and one excluding the exogenous variable. However, in this case the coefficients and IRF of all endogenous variables will change so will not allow a 'true' calculation of the impact of the exogenous variable. Unless I have misunderstood your suggestion.

Thanks,

Rob

EViews Glenn
EViews Developer
Posts: 2671
Joined: Wed Oct 15, 2008 9:17 am

Re: VAR Impulse Response Function and exogenous variables

Postby EViews Glenn » Wed Aug 18, 2021 5:46 pm

Rob,

I *believe* that there was some misunderstanding around Gareth's suggestion (not positive that I am understanding the discussion correctly, but at least that's my sense). Let me see if I can clarify with respect to IRFs and what you appear to want.

In general, from the point-of-view of impulse response functions there is no notion of exogenous or endogenous variable impulses, there is only the notion of impulses derived from shocks to the endogenous variables. In the usual case of, say Cholesky factored impulses, we form three orthogonal impulses from linear combinations of shocks. In this context, there is a specific reason for this choice, but in general impulses can be defined in whatever way we want; we just would like to be able to interpret their meaning.

Now to your specific problem. As you have noted, exogenous variables are not treated like endogenous variables in the VAR and aren't considered in the construction of impulse responses. Interestingly enough, this exclusion actually makes shocks to exogenous variables easier to deal with than shocks to the endogenous variables as the former are basically equivalent to innovations which are easy to interpret (i.e., we don't have to worry about interpreting each impulse as associated with a specific endogenous variable, which is difficult to do in a reduced form with residual correlation).

If I understand what you want correctly, the thought experiment behind an exogenous variable innovations would seem to be "what happens to the endogenous variables in the system when I change my exogenous variable variable from A to B (in your case 0 to 1)"? For the j-th equation, this change is akin to an innovation change of (B-A)*c_j where the c_j is the coefficient on the exogenous variable in the j-th equation. Bear in mind that these changes occur simultaneously since an exogenous variable impact happens in all equations. So ​the relevant impulse is a vector of the coefficients times the change in variable being modeled. In the 0-1 case, the equivalent vector of shocks is simply the vector of coefficients of the exogenous variables.

Conveniently EViews allows you to specify user-defined impulses. The documentation describes how to do so. In your case, create a K x 1 matrix in the workfile (EViews wants a matrix in this case instead of a vector) filled with the coefficients of the exogenous variable of interest in the VAR, and tell EViews to use the matrix in the impulse response routine. If there were a more general exogenous variable change being modeled, simply multiply the matrix/vector by the change of interest.

I hope that the above makes sense, and more importantly, that I am interpreting what you want correctly. If not, we'll try again.

Jawadzureikat
Posts: 1
Joined: Thu Apr 14, 2022 2:17 am

Re: VAR Impulse Response Function and exogenous variables

Postby Jawadzureikat » Fri Apr 15, 2022 9:46 am

Hi there! I am using Eviews 10, and I am trying to test the credit channel mechanism. So, I started with basic Var and then added some restrictions, given that I have got 9 variables 3 of them are exogenous, but, when I created the Var model, I had to add exogenous variables in the box they are assigned to. when I started the impulse response function
it didn't recognize my exogenous variables given that I am interested to know the effect of shocks generated from exogenous variables on the whole system. I would appreciate it if there is any recommendation in this regard?


Return to “Estimation”

Who is online

Users browsing this forum: No registered users and 33 guests