Interpreting (normalized) dummies in explained component oaxaca-blinder

[rachelle dubois]

I am looking at the health gap between indigenous and non-indigenous Canadians and am unsure on how to specify and interpret dummy variables. 2 categorical variables I include are occupation status (employed, unemployed, stay at home, retired) and income group (0-$19,999, $20,000-$39,999, $40,000-$59,999, $60,000-$79,999, $80,000 or more). I see these options:

A) oaxaca healthindex occ_* inc_* [aw=weight], by(indigenousstatus) relax
OR
B) oaxaca healthindex normalize(occ_*) normalize(inc_*) [aw=weight], by(indigenousstatus) relax

So my questions are:
1) Should I normalize these variables?
2) How do I interpret the endowments/explained coefficients for these dummies with or without normalize?

My idea would be, if for example the occ_employed coefficient is .1248:
- A) The health gap would decrease by 0.1248 if indigenous people had the same ratio of employed people as compared to the reference category as non-indigenous people.
- B) The health gap would decrease by 0.1248 if indigenous people had the same rate of employment as non-indigenous people.
Does that sound correct?

Thank you very much!!

[Dung Le]

Hi Rachelle,

I am not sure whether you are using Oaxaca-Blinder decomposition analysis for a linear or non-linear model because you did not mention whether healthindex is a binary or continuous variable. Based on your example codes, I guess you are using the OB for a linear model.

For your 1st question, it is hard to answer because it depends on your research, you may want to read this article for further discussion on normalization of categorical variables: https://onlinelibrary.wiley.com/doi/...1093/ei/cbi053

As for your 2nd question, why don't you show us your results so that people may help you out with interpretation? When you do, please use dataex to provide an example of your data set and use # in the toolbar to show your results. See https://www.statalist.org/forums/help for more details.

For your reference, if you are using a linear model then you may want to read this article: https://journals.sagepub.com/doi/10....867X0800800401
On the other hand, if a nonlinear model is your choice then this article may help: https://papers.ssrn.com/sol3/papers....ract_id=497302

Hope this helps.

Dung Le

[rachelle dubois]

Hi Rachelle,

I am not sure whether you are using Oaxaca-Blinder decomposition analysis for a linear or non-linear model because you did not mention whether healthindex is a binary or continuous variable. Based on your example codes, I guess you are using the OB for a linear model.

For your 1st question, it is hard to answer because it depends on your research, you may want to read this article for further discussion on normalization of categorical variables: https://onlinelibrary.wiley.com/doi/...1093/ei/cbi053

As for your 2nd question, why don't you show us your results so that people may help you out with interpretation? When you do, please use dataex to provide an example of your data set and use # in the toolbar to show your results. See https://www.statalist.org/forums/help for more details.

For your reference, if you are using a linear model then you may want to read this article: https://journals.sagepub.com/doi/10....867X0800800401
On the other hand, if a nonlinear model is your choice then this article may help: https://papers.ssrn.com/sol3/papers....ract_id=497302

Hope this helps.

Dung Le

Hi Dung Le,

Thank you for your detailed reply! I have read both Myeong-Su Yun's and Ben Jann's and I believe I need to normalize my dummy variables and interpret their coefficients in the same way as a regular coefficient.