Mean group estimator for panel VAR using system object

[n1n1ve]

Dear all,

I want to estimate a panel VAR using a mean group estimator in the spirit of Pesaran and Smith (1995). As I understand that Eviews does not have a built-in function for that, I’m considering the following approach and would love to know if it makes sense (or if there are easier ways).

For ease of exposition assume we have 2 countries (let’s call them A and B) in 2 variables (let’s call them X and Y) in a VAR(1). My suggestion is to set up the following system object:

X_A= C(1) + C(2)X_A(-1) + C(3)Y_A(-1)
Y_A= C(4) + C(5)X_A(-1) + C(6)Y_A(-1)
X_B= C(7) + C(8)X_B(-1) + C(9)Y_B(-1)
Y_A= C(10) + C(11)X_B(-1) + C(12)Y_B(-1)

I would then estimate the system as SUR (to take into account potential unobserved factors not included in the model but affecting both countries). The mean group estimator of the coefficient on lagged X in the VAR equation for X is then the mean of C(2) and C(8), the one for the coefficient on lagged Y in the same equation is the mean of C(3) and C(9) etc.

Is this a valid approach? Is there perhaps a more efficient way to set up the estimation (since I have of course more countries, variables and lags in the actual model)?

Thank you very much in advance for your feedback.