Re: st: command for non linear decomposition using a multinimial logit

[Stas Kolenikov <skolenik@gmail.com>]

On Wed, Feb 22, 2012 at 5:45 AM, arlette simo fotso
<fotsoarlette@gmail.com> wrote:
> Can any one please tell me if it exists a Stata commands for doing an
> Oaxaca-Blinder decomposition using a multinomial logit (or multinomil
> probit) model?

As Maarten said, this is not at all straightforward. One reason is
that the binary dependent variable estimate the ratio beta/sigma,
where sigma has to take all of the unaccounted variability, and they
do so by arbitrarily fixing sigma. Thus if one of the groups has a
greater variability in the error term, the estimated coefficients will
not be comparable between two groups. Also, the scale of the
coefficients may change as you add regressors with good explanatory
power (which would in turn decrease the residual variance).

Some additional Stata-related materials on O-B decomposition can be
found at http://ideas.repec.org/p/ets/wpaper/5.html and
http://econpapers.repec.org/paper/bocdsug07/04.htm.

-- 
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/