Oaxaca decomposition

[john smithh]

Hi I’m trying to do a oaxaca decomposition for multiple (more than 2 ) races for my regression. currently I narrowed down my “ethnic” variable into just 2 variables but that would mask a lot of interpretation. can someone teach me the codes to allowing multiple groups of ethnics?


Currently my code is

xi: oaxaca lnhourpay sex age educ industry married workerlevel employmentduration famunit publicsector regsmoke region fulltime , by(white) pooled noisily

Looking to change “white” which is 2 categories: white and non white to ethnics which have multiple categories: white chinese bangladesh indian etc..


Thanks for you help in advance!

[FernandoRios]

Hi John,
The short answer is that what you are asking is not possible. Oaxaca Decomposition is made to compare only two groups at a time, which is by construction what you need to obtain a gap. There is, however, no limitation of how many "pairs" you can use to analyze the data.
Perhaps an strategy that can be used here is to compare the overall average wage equation (a regression with the pooled information of all ethnicities), and compare that common reference group with each individual ethnic wage equation separately. That way, at least your reference group would be consistent.
HTH
Fernando

[john smithh]

Hi fernando! Thanks for the response I just tired what you mentioned but I’m having a hard time with the coding :x Would you mind maybe typing the oaxaca regression you mentioned?

Just to give bit more detail, my “white” variable consist of white and non white
my ethnic variable consist of White Black Chinese Pakistani Bangladesh. So how do i compare using pooled information? Do i put in ethnic in my regression and “by” it with the white variables ?