# st: RE: Panel data simultaneous equations

 From "Gielis Wim" <[email protected]> To <[email protected]> Subject st: RE: Panel data simultaneous equations Date Wed, 20 Jul 2005 00:21:28 +0200

```[Apologies if the mail was sent twice. Small mail problem...]

Hello all,

Would it perhaps be possible to take up dummy variables for year and subject effects (but beware for perfect collinearity)

Actually, I face a similar problem. The setup is the same as the one of Tewodaj, but in addition:

- 1 of my (2) equations has nonlinearities in the parameters
- I need an IV-method because of endogeneity issues and lagged dependent variables

So, my questions to the other Statalisters,

1. is the proposal advice correct?
2. if yes, how to go estimating this in Stata?
3. any suggestions for my (extended) problem?

Best regards,

Wim Gielis
University of Antwerp
Belgium

-----Oorspronkelijk bericht-----
Van: [email protected] namens TEWODAJ MOGUES
Verzonden: di 19-7-2005 23:54
Aan: Stata _
Onderwerp: st: Panel data simultaneous equations

Hello List,

I did an extensive search on statalist before posting this question,
and unfortunately I didn't find a solution as yet, so I thought I'll
try posting myself.

I want to estimate a simultaneous equations model in which each
structural equation has a panel data structure. Here is an example of
the type of system I mean:

y1_it = beta0_i  + beta  * X1_it + u1_it
y2_it = gamma0_i + gamma * X2_it + u2_it
.
.
.
yG_it = theta0_i + theta * XG_it + uG_it

I.e. there are G structural equations, each having T number of time
periods. So yg_it is a scalar and Xg_it is a K_g x 1 vector, and this
vector may include yh_it (where h refers to some structural equation
different from g). So I want to *jointly* estimate beta,
gamma,...,theta, say by 3SLS, just like I would in the more
conventional cross-section system:

y1_i = beta  * X1_i + u1_i
y2_i = gamma * X2_i + u2_i
.
.
yG_i = theta * XG_i + uG_i

in which yg_i is a scalar. The latter I would estimate, say, using
reg3, that's straightforward. And estimating a single-equation panel
equation such as

y_it = beta  * X_it + u_it

is also straightforward, e.g. using xtreg, xtivreg etc. depending on
the particular estimation desired. However, when I have a *combination*
of panel data and system of equations I am at a loss how best to
approach this in STATA. Doing this equation by equation of course
defeats the purpose of estimating as a system.

Has anyone else faced this, and if (as is my impression) there is no
existing command for this, is there anyone out there who has written a
stata programme for estimating a panel simultaneous equations model of
this sort?

Thanks,
Tewodaj
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```

<<winmail.dat>>