Blanchard and Perotti SVAR model

[dalida]

Hi,
I want to estimate the impact of fiscal shocks on GDP using the structural VAR approach applied by Blanchard and Perotti.
In first stage I estimate all parameters of A and B matrices verifying A ut = B vt where ut represents the residuals of standard VAR and vt is the vectot of structural residuals.
My question is after estimating all parameters of A and B using immediate elasticities and instrumental variables regression (the Blanchard and Perotti approach) how can I obtain yhe response of endogenous variables (taxes, expenditure, gdp, inflation, interest rate) to a structural shock of public expenditure?

[EViews Matt]

Hello,

So you have a standard VAR estimated and your custom A and B matrices, correct? If so, you can form the short-term response matrix S from A and B, extract the structural shock you're interested in from S, and then use that custom shock with the VAR object's impulse response procedure. Programmatically, those three steps are:

Code: Select all

matrix S = @inverse(A) * B
vector shock = @columnextract(S, 2)   'Extracting the second structural shock.
yourvar.impulse(imp=user, fname=shock)


The last step can be performed via the GUI. Within the VAR impulse responses dialog, on the impulse definition tab, select the user specified decomposition method and enter then name of the vector object containing the custom shock, e.g., "shock" in my example.

[dalida]

Thank you for your reply. But my problem lies in estimating the parameters of matrices A and B. According to the Blanchard and Perotti approach, the matrix form of the model to be estimated is presented in the attachement

parameters.docx

(36.34 KiB) Downloaded 241 times

Where the terms α correspond to elasticity; the terms β correspond to the cross reactions of the variables of the public finances, and the terms ϒ are the other coefficients to be estimated using instrumental variables regression.
My problem is how to estimate these ϒ parameters?
The Blanchard and Perotti approach consists on :
1/ we suppose that βtag = 0 and estimate by OLS the second equation
2/ we use the estimated residuals of the first and second equations to estimate the third equation by the TSLS – Two stage least squares method.
3/ we use the estimated residuals of the third and fourth equations to estimate the last one by TSLS method.
When I proceed with this approache I obtain all coefficients of A and B matrices, so we haven’t unknown coefficients to estimate with SVAR model and Eviews don’t accept the B matrix.

[EViews Matt]

I'm a bit confused by your follow up. Have you successfully estimated all the coefficients in the A and B matrices?