System Estimation by 3SLS using Panel Data

[IEichler]

Hi,

I am trying to estimate a four equation system by 3sls using panel data. Is this possible in E-views? So when specifying a system and having set up the data as panel data, does the estimation take this into account?

Thanks a lot!

[EViews Gareth]

Depends upon what you mean by "take this into account".

When you estimate the system, EViews knows that you are sitting on panel data, so any lags/leads/dynamics you have in your equation will take the panel structure into account, and lags/leads will not cross-seams (i.e. they won't go over cross-sections).

However there is no special panel based system estimation techniques - you can not, for example, estimate fixed (well other than manually with dummy variables) or random effects.

[demi]

Hey,
I'm also trying to estimate a three equation system by 3sls using panel data.
I'm a lttle bit confused now how to actually estimate that in eViews via 3sls... I thought about using Two-Stage-Least-Square and SUR afterwards. Is that correct? Or would you go for the GMM estimation? Are there any codes you could recommend?

Thank you. Any input is highly appreciated...!

[EViews Gareth]

Not sure I follow the question. Why not just estimate by 3SLS?

[demi]

hm, ok. But how would that work? No clue how to compute that. Any help?

Thank you.

[EViews Gareth]

The same way you would estimate by 2SLS or SUR, but change the method to 3SLS instead...

[demi]

hm, I somehow can just chose from LS, TSLS, GMM, BINARY, Ordered, Censored, Count, Qreg, StepLS. So there is no 3SLS. What am I doing wrong???

Thank you

[EViews Gareth]

You need to be using a system object, not an equation object.

[pablo_eviews]

Depends upon what you mean by "take this into account".

When you estimate the system, EViews knows that you are sitting on panel data, so any lags/leads/dynamics you have in your equation will take the panel structure into account, and lags/leads will not cross-seams (i.e. they won't go over cross-sections).

However there is no special panel based system estimation techniques - you can not, for example, estimate fixed (well other than manually with dummy variables) or random effects.

Hello,

I'm having some trouble while estimating a system with panel data. How should I introduce both period and cross-section dummies into the spec tab? There are no preset options for that (period and cross-section dummies that is) in the system window under panel data. Should I construct them manually? I have a really large panel!

Thanks for the help in advance.

Pablo

[EViews Gareth]

EViews doesn't currently support built in cross/period fixed/random effects. You'll have to create the dummies manually.