st: Re: probit and ordered probit eqns in a simultaneous system

[Michael Alexander <Michael.Alexander@aifs.gov.au>]

Hi

I posted this query to the list last week but unfortunately noone has 
had the opportunity to respond as yet. I am posting it again in case 
someone is now able to help. Or if my query is an inappropriate one for 
the list, perhaps someone can let me know.

Regards
Michael

Michael Alexander wrote:


> Hi listreaders
>
> I have a particular econometric problem that I am trying to analyse in 
> Stata. I am trying to model employment of mothers (binary dependent 
> variable) with the inclusion of self-reported health as an explanatory 
> variable. I am postulating that self-reported health (a five-level 
> categorical variable) is endogeous to employment. Following Cai and 
> Kalb (2006) I am also postulating that labour force status is also a 
> determinant of self-reported health (although this is of secondary 
> importance to me).
>
> Consequently, my modelling approach was to do a system of two 
> equations (one for LFS and one for self-reported health) and estimate 
> them simultaneously. Unfortunately, I have only found the -mvprobit- 
> command which will sort of allow me to model this, but only if I 
> convert my self-reported health variable to a binary variable. What I 
> ideally need is a command that will allow me to simultaneously model  
> a binary dependent variable (employment) and an ordinal categorical 
> dependent variable (self-reported health, which has five-levels). Does 
> such a command exist? Or am I able to build my own estimator ? (Not 
> something I am keen on as I haven't done a lot of programming but am 
> willing to give it a go if this is the only approach).
>
> An alternative approach would be to not worry about the impact of LFS 
> on self-reported health and only focus on the endogenous nature of 
> self-reported health on employment. That way I could use an 
> instrumental variable approach to dealing with the endogeniety of 
> self-reported health in an employment equation. But as far as I can 
> make out, this still requires me to model self-reported health as a 
> binary variable (using -ivprobit-). Can I use an instrumental variable 
> approach even when my endogeous variable is (ordinal) categorical.
>
> Thanks in anticipation
> Michael
>
> Cai, L. and Kalb, K. (2006), 'Health Status and Labour Force 
> Participation: Evidence from the HILDA Data', Health Economics, vol. 
> 15, March, pp. 241-261.
--
Michael Alexander
Principal Research Fellow
Australian Institute of Family Studies

300 Queen Street 
Melbourne VIC 3000

Tel: 03 9214 7841
Fax: 03 9214 7839
Mob: 0419 406 078
Email: Michael.Alexander@aifs.gov.au
Web-site: www.aifs.gov.au

*
*   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/