[Help] Three stages least squares method

[yohoyihi]

I plan to use Three stages least squares estimation method for my thesis project, is this method available on Eviews 6 ? As i could not found any three stages least squares command on eviews (sorry i'm newbie ). My thesis is about simultaneous equations, as i have 4 equations on these simultaneous equations model.

Would you mind to give me some advices (step by step) about doing three stages least squares estimation on eviews 6 ? I really need help from some eviews experts here

Regards
Tony

[startz]

I plan to use Three stages least squares estimation method for my thesis project, is this method available on Eviews 6 ? As i could not found any three stages least squares command on eviews (sorry i'm newbie ). My thesis is about simultaneous equations, as i have 4 equations on these simultaneous equations model.

Would you mind to give me some advices (step by step) about doing three stages least squares estimation on eviews 6 ? I really need help from some eviews experts here

Regards
Tony

Create a system object, enter your equations, and choose 3sls as the estimation method.

[yohoyihi]

simultaneous models from literature are:

Gr(Y) = a0 + a1(I/Y) + a2(FDI/Y) + a3GrL +a4Gr(X) + t
FDI/Y = b0 + b1Gr(Y) + b2 (Y/N) + b3Gr(Pw) + t
I/Y = c0 + c1Gr(Y/N) + c2(Y/N) + c3Gr(X) + c4Gr(Kf) + t
Gr(X) = d0 + d1Gr(Y) + d2REER + d3(Y/N)+d4Gr(Yf) + t

a0, b0, c0, d0 are constant number
t is an error terms

how do i transform and enter these equations into eviews correctly ? I've generated all variables and data that i will use into eviews, and give a try to do 3sls. But I got an error message that a0 is not defined. Did I miss something here ?

regards,
Tony

[startz]

simultaneous models from literature are:

Gr(Y) = a0 + a1(I/Y) + a2(FDI/Y) + a3GrL +a4Gr(X) + t
FDI/Y = b0 + b1Gr(Y) + b2 (Y/N) + b3Gr(Pw) + t
I/Y = c0 + c1Gr(Y/N) + c2(Y/N) + c3Gr(X) + c4Gr(Kf) + t
Gr(X) = d0 + d1Gr(Y) + d2REER + d3(Y/N)+d4Gr(Yf) + t

a0, b0, c0, d0 are constant number
t is an error terms

how do i transform and enter these equations into eviews correctly ? I've generated all variables and data that i will use into eviews, and give a try to do 3sls. But I got an error message that a0 is not defined. Did I miss something here ?

regards,
Tony

You need to use a coef statement to tell EViews that a0, etc. are coefficients, or use the predefined coefficients, c(1), c(2), c(3)...
(Also, remember to use "*" for multiplication.)

[yohoyihi]

I used the predefined coefficients such as c(1), c(2), c(3), ....

i tried to estimate the equations system by using 3sls, unfortunately i got an error message that said there were no instrument list specified. How do I specify the instrument list on eviews before I regress equations ? thanks.

regards,
Tony

[EViews Gareth]

You can either include a line:

@inst z1 z2 z3 etc....

where z1.. are the instruments for the whole system, or for equation specific instruments, you can include them after an @ in the equation specification:

y=c(1)+c(2)*x1 @ z1

[yohoyihi]

I've been working with eviews for days, and using 3sls. Unfortunately, after I have regressed my simultaneous systems, I got some strange results from eviews Output. I found that My all system's R-square was minus ?? geez, Did I missed something here ? I've been also tested stationarity each variables as time series models, and entered as first or second difference if some variables had problem with stationarity.

My question is, did I missed something important processes here ? I think I should post my 3SLS syntax so you guys can help me

*this is my standard 3sls system's syntax without any adjustment
@inst ipery grl grw reer gryworld
gry = c(1) + c(2)*ipery + c(3)*fdipery + c(4)*grl + c(5)*grx
fdipery = c(6) + c(7)*gry + c(8)*pdbkapita + c(9)*grw
ipery = c(10) + c(11)*grpdbkapita + c(12)*pdbkapita + c(13)*grx + c(14)*grir
grx = c(15) + c(16)*gry + c(17)*reer + c(18)*pdbkapita + c(19)*gryworld

The funny idea here, I've tried with other estimation methods (SURE and OLS), surprisingly i got positif R-square although they are not in a good number (around 20% - 5%)

Did I missed something guys ? I Really need big help here

Thanks,
Tony

[startz]

I've been working with eviews for days, and using 3sls. Unfortunately, after I have regressed my simultaneous systems, I got some strange results from eviews Output. I found that My all system's R-square was minus ?? geez, Did I missed something here ? I've been also tested stationarity each variables as time series models, and entered as first or second difference if some variables had problem with stationarity.

There's nothing necessarily wrong with a negative R-square when using any instrumental variable method, including 3SLS. R-square tells how well you fit the data. That's maximized with OLS. The purpose of 3SLS is to get accurate estimates of the coefficients.

[yohoyihi]

Here you are, my workfile. I hope someone can give me some advices after examined this so I can correct my mistakes

[yohoyihi]

here you are my simultan dataset in xls file

[EViews Gareth]

I didn't notice any mistakes. What makes you think you have mistakes?

[yohoyihi]

I don't know what's wrong here, Gareth. I just want to replicate some previous research in this topic (which is have a positif and reasonable R Square). I feel it is "uncomfortable" to explain your econometrics model and output with MINUS R Square.

Could someone provide me a good reason or some arguments that in simultaneous and Instruments variables method (maybe website links) Minus R Square is acceptable ?

Thanks,
Tony

[EViews Gareth]

A simple Google search of "negative r square" will give you lots of results.