Am 23.04.2020 um 15:31 schrieb Olasehinde Timmy: > I am finding it difficult to use the additional long-run restriction in > the SVEC interface. For instance, a four-variable VEC model with one > cointegrating vector requires additional 3 long-run restrictions with > the 3 linear independent shocks to exactly identified the system. To avoid misunderstandings: Do you absolutely want to avoid short-run restrictions? Because it is not necessary to add further _long-run_ restrictions, instead that depends on the application and the economic logic. Achieving identification through only long-run id may be difficult. > 1) Is it possible to use the total SVEC gui to do the restrictions ? In principle, yes. (When you formulate your model and the cointegration setup, remember that the equations associated with the permanent shocks must come first, by convention.) Secondly, please read footnote 20 in the SVAR manual. And finally, you can formulate additional long-run restrictions with a pattern matrix that you then paste (its name I mean) into the field saying "Restr pattern (further long-run C)". The structure of this pattern matrix is just like the one that specifies short-run restrictions. If you have further questions about this, it would be helpful to have an idea of the concrete restrictions that you want to impose, both short-run and long-run. cheers sven _______________________________________________ Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/

Am 23.04.20 um 22:44 schrieb Olasehinde Timmy: > I want to estimate a five-variable SVECs model with; > Y = Output > T = Tax > R = Interest rate > G = Government expenditure > P = Price > I assumed three theoretical co-integrating vectors; > Y = beta11*G + beta12*P (stabilization rule) > T = beta21*G (solvency) > R = beta31*P (fisher relation) > > I want to however assign the three transitory shocks to T, G, & R > respectively; does this not violate the listing the transitory shocks > last in the Gretl convention? OK, it's very good that now we have a specific example to work with. First, you're right that the variable ordering matters here, so you would have to construct your variable list for example like this: Y P T G R. However, since you want Y and P to appear in the same cointegration relationship, the default normalization for the beta matrix in the VECM then won't work here (because it puts an identity submatrix at the top). So in this case you need to estimate your beta explicitly with your cointegration restrictions imposed before you go to the SVEC. So: You have to use gretl's builtin VECM apparatus and specify your (over-) identifying restrictions on beta. (After initial estimation go to Tests/Linear restrictions, then you have to type in the restrictions in the format explained in the gretl user guide.) To be on the safe side you should explicitly "grab" the estimated beta matrix: In the main window go to Add/Define matrix and type something like "mybeta = $jbeta". OK, so now you can enter the estimated cointegration matrix into the SVEC_GUI window field "coint (j)beta". Either you simply type in gretl's accessor name "$jbeta" (without the quotes) or if that doesn't work for some reason you use your grabbed object "mybeta". Then you can proceed to specify the remaining structural shock restrictions with the pattern matrices. HTH sven _______________________________________________ Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/