Todd,
Have looked at heckprob (for probit estimation with selection) and combining it
with suest - seeming unrelated estimation) from STB -52.
I hope this helps,
Scott
----- Original Message -----
From: <[email protected]>
To: <[email protected]>
Sent: Wednesday, January 15, 2003 3:19 PM
Subject: st: commands/models for handling dummy endogenous variables
> Dear Statalist,
>
> I am developing a model to estimate the effect of managed care on the types
> of services offered at drug abuse treatment facilities.
>
> Managed care (MC) enters the model as a dummy (either a given treatment
> facility has a relationship with a managed care organization or it does
> not).
>
> Each service offering (SO) also enters the model as a dummy (either the
> treatment facility offers the service or it does not).
>
> All of the data are from year 2000.
>
> A very naive approach would use a one equation probit such as:
>
> SO = MC + X + e
>
> where SO = a specific service offering dummy, MC = managed care
> dummy, X = rest of the regressors, e = error term.
>
> However, there is most likely either selection or endogeneity (or both?)
> between SO and MC. That is, managed care may enter a facility and, with an
> eye on the bottom line, gradually alter the facility's service offerings
> (this is what I'm really trying to measure). Alternatively, managed care
> may seek out facilities that already have a certain suite (or portfolio) of
> service offerings. Or, facilities themselves may alter their offerings in
> hopes of attracting managed care patients.
>
> In short, I don't think it's appropriate to interpret the coefficient of MC
> in the above one-equation probit as the "effect of MC on the likelihood
> that a particular service will be offered by a drug abuse treatment
> facility" (am I wrong about this?).
>
> What I'd like to do (I think), is to run a 2 equation model such as:
>
> SO = MC + X + e1
> MC = SO + Z + e2
> where Z probably overlaps with X to a large extent.
>
> Although -biprobit- or -sureg- will take the data and run the numbers, I am
> told that both are inappropriate because they ignore the endogeneity issue.
>
> By any chance, does anybody know whether there is user-written code that
> will handle the above "simultaneous equations probit model?"
>
> Sincerely,
> Todd Olmstead
*
* 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/
