Jann (2008) The Stata Journal 8(4): 453-479 Question

[Tim Sulls]

I was reading Jann's (2008 The Stata Journal 8(4): 453-479) insightful article and on page 15 and 16 they discuss what endowment, coefficient, and interaction represent in an analysis. I started playing with the data and have some questions that I hope the community can answer.

use http://fmwww.bc.edu/RePEc/bocode/o/oaxaca.dta
oaxaca lnwage educ exper tenure, by(female) noisily

1a) Does the endowment effect mean how women would do (wages) if they had similar mean values of predictors as men?
1b) If there another way to explain this more clearly?
1c) When I run the model myself I see that under endowment section that educ is significant but tenure is not. How would we interpret at the predictor level, for example educ?
1d) Would we say that tenure doesnt matter?

2a) How is the value of the coefficients section (0.0825) interpreted?
2b) All the article states is that it is often discussed as "discrimination". What does this mean?
2c) How do we interpret independent coefficients in this area? Imagine tenure is significant but the rest are not. What would that mean?

3a) They do not discuss how to interpret interaction. What does this section mean?
3b) Imagine that educ is significant, how would we interpret?

[Carlo Lazzaro]

Tim:
I would email Jann directly, as he is the (corresponding) author of that community-contributed module.

[Stephen Jenkins]

Rather than write to Ben, I'd suggest reading some of the source articles cited in the help-file about this decomposition-of-means method (attributed to Blinder, Oaxaca and sometimes Kitigawa -- but see also https://docs.iza.org/dp16188.pdf). My point is that the OP's questions are about the method and its interpretation, not the code, and there is a huge literature about this..

[Carlo Lazzaro]

Stephen:
reading the post once more, I think you're right.

[Tim Sulls]

I did e-mail directly a while back and never heard. I have also read a lot of the papers but it gets technical very fast and I don't understand. I was hoping an explanation of these simple effects would then allow me to best read the technical. But if people here do not know, I feel better about it.

[Andrew Musau]

Note that responding to posts here is a voluntary service; no one is paid. A lot of us do OB decompositions, so it is unlikely that lack of knowledge explains why your questions were not answered. If you search this forum, you will find discussions related to interpreting output from oaxaca. However, questions such as "What does it mean to have an insignificant coefficient?" are too basic to have been discussed. Ultimately, if you have read the recommended papers and still cannot answer the questions you posed, and assuming no one responds, approach your supervisor—who is paid to guide you. If you are not a student, consider finding a consultant; this may involve paying for the service.

[Tim Sulls]

Thanks, Andrew. I am not a student and do not have funds for a consultant. I was hoping for some answers and understanding that this is voluntary. I searched the board but could not find answers. Maybe because too basic. But I guess I hoped that the basic level would be an easy answer for someone if possible.

[Andrew Musau]

See https://www.statalist.org/forums/for...-decomposition for one such discussion relating to interpretation of results from the command. There are more, use keyword "oaxaca" to search. On the reference to discrimination, the original applications focused on decomposing differences in the wages of men and women (gender pay gap), or black and white (racial pay gap). The Wikipedia entry of Blinder-Oaxaca decomposition is not that technical and discusses all these matters: https://en.wikipedia.org/wiki/Blinde..._decomposition.

[Tim Sulls]

Hi Andrew. I appreciate you searching for me. These are links I have found. The problem with the initial Statalist discussion you post is that the very informed Wei says that they do not know how to interpret the 3 way which is what I am asking above. There is also only loose discussion of the unexplained per the above questions. I also appreciate the Wikipedia article but having read it last week it also does not answer the majority of questions. The one insight is that unexplained isnt just discrimination but it doesnt explain what discrimination means. Is it that tenure for women is rewarded less than men (unexplained) where explained means women may have less of it?

I have looked a lot before posting and I guess I was hoping someone would take the time like they did in these posts to help. I will continue reading and checking back in hopes.

[Andrew Musau]

The thread is open. Anyone who is willing to help will do that.