Dear Statalist,
The Stata-provided function IV Tobit appears to estimate a model in
which the ultimate dependent variable is censored but the instrumented
variable is not. Is there a Stata routine to do the opposite --
estimate a model in which the ultimate dependent variable is
uncensored, but the problematic endogenous variable that needs to be
instrumented *is* censored?
A simple example in structural form:
Wage = B0 + B1 Time_in_training_sessions + B2 Experience + error_term
Time_in_training_sessions = G0 + G1 Wage + G2
Dislike_for_training_sessions + error_term if G0 + G1 Wage + G2
Dislike_for_training_sessions + error_term > 0 and 0 otherwise
Here, Dislike_for_training_sessions is an instrument for
Time_in_training_sessions, but it takes a value of 0 for many
individuals.
The online and paper documentation makes it clear that IV Tobit does
not do what I want (it only handels censoring in the primary
regression, not in the instrument regression).
The amazingly flexible cmp seems to require a diagonal structure,
precluding the mutual endogenaity likely in IV cases.
Am I overlooking a way to use cmp, does another procedure exist, or do
I need to work from scratch? I don't want to re-invent the wheel if
this is out there somewhere.
Peace,
Daniel Lawson
Assistant Professor of Economics
Drew University
[email protected]
*
* 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/
