Two days ago there has been a query about standard
errors for the components of the (Oaxaca/Blinder)
decomposition of wage differentials. Stas answered:
> As long as all those decompositions are linear functions of the
> coefficients (as far as I can recall), you can get the standard errors
> from -lincom- command, although you would have to do all of the
> algebra from scratch. If there are any non-linearities involved, the
> goal can be achieved with -nlcom- instead.
There are two problems with -lincom- or -nlcom-. Firstly,
the decomposition of wage differentials is usually
preformed using two sets of estimates (e.g. for male
and for female), but -lincom/nlcom- can only be applied
to one set of estimates (there might be a trick, however,
to overcome this problem).
Secondly and more seriously, the standard errors
computed by -lincom- and -nlcom- would only be valid,
if the regressors are fixed (this is also true for the
formulas in Oaxaca/Ransom 1994), because the means of
the regressors, which have nonzero variance in the case
of stochastic regressors, are used in the decomposition.
I recently wrote down some formulas to take into account
stochastic regressors. Unfortunately, the new version
of -decompose-, which computes these standard errors
(as well as the standard errors for the case of fixed
regressors) is not finished yet. People interested in
the formulas and the beta-version of the program should
contact me offlist.
This brings me to the next point: Stas complained that
"decompose" is not a good name for this package. I will
be happy to use another name for the new version if
anyone makes a good suggestion.
Cheers, ben
*
* 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/
