# st: Interpretation of Oaxaca-decomposition

 From Jochen Mistelbacher <[email protected]> To [email protected] Subject st: Interpretation of Oaxaca-decomposition Date Wed, 12 Dec 2007 16:40:43 +0100

Dear statalist members,

I would like to consult the list on some basic, partly non-technical questions regarding the Oaxaca-decompostion, or regression decomposition.

My setting: I have 2 regions (factor1) with 5 groups (factor2) within each region. I run normal regressions including dummies etc, and now would like to decompose the differences of the mean depvar between different groups, e.g. group1-group5 in region=1, group1-group5 in region=2, etc., being esp. interested in the explained part of the decomposition (differences in means of the independent variables).
Using Ben Jann`s "oaxaca"-routine with the overall pooled regression as the reference I get e.g. the following basic results: The explained portion of the difference in the mean of depvar in region 1 is ca. 85-90%, in region 2 it is around 110%. Now,

1) is it correct to state that in region 1 the difference in the mean depvar is higher than the mean differences in the independent variables would suggest (given the included variables), whereas in region 2 the mean difference of the depvar should be higher, thus that in region 1 there seem to be factors / variables at work which lead to a larger difference, whereas in region 2 there seem to be factors working in the opposite direction, given the included independent variables?

2) Esp. in region 2 there is some problem regarding between-group heteroscedasticity. Is there a way to solve this, or would it be correct to e.g. use a WLS-regression with groupspecific analytical weights as a pooled reference for the regression composition?

Thanx in advance for comments and suggestions,

Jochen
*
* 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/