EViews.com

Dear Eviews community,

I have a problem with the specification of a simultaneous equations model which I want to estimate via GMM (I am replicating a model):

Specifically, I have two equations:

1) Y1 = c(0) + c(1)*Y2 + c(2)*X1 + c(3)*X2 + c(4)*X3+ c(5)*X4
2) Y2 = c(6) + c(7)*Y1 + c(8)*X1 + c(9)*X2 + c(10)*X3 + c(11)*X4 + c(12)*Z1 + c(13)*Z2 + c(14)*Z3

- Y1 & Y2 are endogenous
- X1-X4 are independent, exogenous variables and are described as "included instruments"
- Z1-Z3 are independent, exogenous variables and are described as "excluded instruments"

I am now having trouble correctly specifying this system in Eviews. One attempt of mine is the following code, it's estimation results however in the error message "Matrix near singluar":

Y1 = c(0) + c(1)*Y2 + c(2)*X1 + c(3)*X2 + c(4)*X3+ c(5)*X4 @ X1 X2 X3 X4
Y2 = c(6) + c(7)*Y1 + c(8)*X1 + c(9)*X2 + c(10)*X3 + c(11)*X4 + c(12)*Z1 + c(13)*Z2 + c(14)*Z3 @ X1 X2 X3 X4 Z1 Z2 Z3

It would be great if you could help me out with this problem! I am sure the solution is fairly obvious but I, as a Eviews beginner, can't see it.

Sincerely, felix

You don't have enough instruments for the second equation.

Hey Startz,

thanks for the quick reply! I am not sure about not having enough variables: I am replicating this model from a working paper (page 10, http://www.frbsf.org/publications/econo ... 0-17bk.pdf ) in which this model was successfully estimated. Since they estimated this model with the variables I have, I think I should be able to do the same.

One thing I forgot to mention about Z1-Z3 is that the are instruments for Y2. There are also no "excluded" instruments for Y1, just the included instruments X1 X2 X3 X4.

It be great if we could solve this problem.

felix

You can't write C(0). C is one based, not zero based. But, Startz is right - you don't have enough instruments.

Hey Gareth, thansk for your reply! Could you be more specific with regards to the instruments I am missing (how many? for what eq?)? Again, since I am just exactly replicating a model used and successfully estimated by a different researcher, I am having difficulties understanding how it can be possible that there are not enough instruments. I must be missing/misunderstanding something, but I don´t know what. I am certain that the model int terms of equations, variables and instruments is correct, do you any suggestions what else I could look at?

Thanks for your advice! felix

The first equation has a constant plus five coefficients. You've only given four instruments.

Following on from Startz - EViews will add a constant as an instrument automatically, so you're at least one instrument short.