restricted_VAR

[lsym]

eviews6: I would like to ask if naybody knows how i can estimate a restricted (3*3) VAR model, where the coefficients of the lagged values of the third variable in the system are restricted to be zero (i saw it in a paper). The system is specified as follows:

X(t)=a1+a(1,1)X(t-1)+b(1,1)Y(t-1)+e(1,t)
Y(t)=a2+a(2,1)X(t-1)+b(2,1)Y(t-1)+e(2,t)
Z(t)=a3+a(3,1)X(t-1)+b(3,1)Y(t-1)+e(3,t)

where Z(t)=X(t)*Y(t) and a(.), b(.) the coefficients of the lagged endogenous variables

Thanks

[EViews Gareth]

You'll have to estimate it in a system object rather than a VAR object. In the system you explicitly write out the relationship between the coefficients and variables, so can impose any restrictions you want.

[lsym]

Thanks very much, actually i was thinking of the same thing but i am not sure if it's correct. To be honest i believe that the results will be different since the system is based on different assumptions. But still it's a good solution as long as the objective of my analysis is not sensitive to the estimation method, i only need in-sample estimates.

Thanks again

[EViews Gareth]

If you estimate an unrestricted VAR in both a VAR object and a system object, you'll notice that the results are identical.

[lsym]

Ok i checked it you are right, thanks!