I am estimating a system of equations simultaneously
determined (3 equations) with panel data using reg3 (3SLS
fixed effects). I=E2=80=99m interested in performing a test for; (i)
overidentifying restrictions and exogeneity; and (ii)
endogeneity. If my equation weren=E2=80=99t simultaneously
determined I wouldn=E2=80=99t have any problem using xtivreg, fe and
performing these tests with overidxt and hausman (equivalent
to ivendog).
>From Stata=E2=80=99s website it reads that =E2=80=9CHausman test in Stata
9.0 can be used to test for exogeneity or overidentifying
restrictions for two-stage least squares (2SLS) and three-
stage least squares (3SLS)=E2=80=9D. I would like to know if I can
perform a test (and how) for; (i) overidentifying
restrictions and exogeneity of the instruments in each of
the equations; and (ii) endogeneity of each of my endogenous
variables if, for example, I=E2=80=99m estimating the following
model:
reg3 (Y1 Y2 Y3 X1=E2=80=A6.Xn) (Y2 Y1 Y3 X1=E2=80=A6.Xn) (Y3 Y1 Y2 X1=E2=80=
=A6Xn)
If it is not with a Hausman test, are there other ways to
perform these tests after reg3?
It would be valid to perform these tests equation-by-
equation using xtivreg and then, if they don=E2=80=99t provide
enough evidence to reject the validity of the specification
and instruments, estimating the full system of equations
with reg3?
Thanks in anticipation,
Felipe
_______________________________________
Felipe Targa
Ph.D. Candidate
Dept of Civil & Environmental Engineering
1173 Glenn Martin Hall
University of Maryland, College Park MD 20742
www.wam.umd.edu/~ftarga
_______________________________________
*
* 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/
