Blinder-Oaxaca Decomposition for a Multivariate Linear Regression

Hello Everyone,

I am working on a thesis that uses a multivariate linear regression to evaluate the racial wealth gap between four groups: whites, blacks, Asians, and Hispanics. If it helps, I am using SPSS complex samples, general linear models, and a large nationally representative survey called the Survey of Income and Program Participation (SIPP).

I have already run four separate regressions for each subgroup. Now I am trying to use something called a "Blinder-Oaxaca" decomposition described here. Basically, it breaks the regression into two components: the explained and the unexplained, and allows me to calculate how much each independent variable contributes to the explained percent. It has been used in a number of publications seen below:

Can anyone help give me some practical, less equation heavy, advice on how to use this method? Or point me towards a guide/source that could? I'm not sure if I could simply use the results of my four regressions and calculate the decomposition by hand, or if I need to somehow plug the intercepts, coefficients, and means back into a program. I'm also generally confused on how to calculate it.

Thank you, and I really appreciate any help!
Hi Colin, I think I have the same problem, I found the way to do it in STATA, and I'm fixing some issues that are important in my case. If you need the program I found and if it will be helpful for you, just let me know.
I think I'm doing a thesis very similar than your's and maybe the paper written by Yun (2004) will help you a little.

All the best,