Dear listers,
I am estimating a model which involves three steps (or actually 2) and
need to correct for the standard errors. I am thinking of the
Murphy-Topel (M-T) correction but would like to know whether it makes
sense and how to go about it in Stata:
Step 1) I estimate a count model (negative binomial)
M = aZ + e
The prediction of M is used as an instrument for R, a reduced form equation.
In steps 2 and 3) , I estimate an IV-Tobit
L = bR + gX + u (structural equation)
R = dY + e (reduced form equation)
Since the prediction of M is my instrument for R, there is a downward
bias in the standard errors of the structural equation that a naïve
variance would not account for. Can an M-T correction of the standard
errors produced in the structural equation solve this problem and if yes
how could I go about it in Stata? Your suggestions are highly appreciated.
Happy holidays!
Evans
begin:vcard
fn:Evans Jadotte
n:Jadotte;Evans
org;quoted-printable:Universitat Aut=C3=B2noma de Barcelona ;Applied Economics
adr:;;Campus de Bellaterra;Bellaterra;Barcelona;08193;Spain
email;internet:[email protected]
tel;work:+34 93 5813415
tel;fax:+34 93 5812292
tel;cell:+34 655319494
version:2.1
end:vcard
