-oaxaca- Gender Wage Gap

[Thea Fredriksen]

Hello,

I am analysing the gender wage gap in the UK using the Oaxaca-Blinder decomposition. I am including control variables for age, education, region etc. In a normal regression, I would add age squared as a control but I am not sure how to include this when using the -oaxaca- command. I will first perform a simple decomposition, only including age as a control in order to explain the intuition behind the method.

I have created a variable for age^2 = age2 and this is the code I am currently using.

oaxaca lnwage age age2, by(female) noisily

Is this correct or is it better to drop the age^2?

[FernandoRios]

That is correct, however keep in mind that interpreting squared parameters in OB decompositions is not easy.
The aggregate decomposition, on the other hand will be better

[Thea Fredriksen]

Thank you!
Yes, I am having some trouble interpreting the numbers, as the age squared is a large negative number and "age" suddenly became very large.

Could you help me with the interpretation of these numbers?

[Thea Fredriksen]

FernandoRios, I realised I forgot to mention you in the previous post.
Thank you.

[FernandoRios]

Hi Thea,
they way I would interpret this:
1. RAW Wage gap is 14.46% (or 0.1446 log points)
2. Age affect = age+age2
Endowment = 0.032008-0.0274666 =0.004
Thus, Differences in age in the sample explain about 0.4% of the total
3. Coefficients:
Age+age2=1.188125-0.5179892 ~~ 0.67
Constant = -.53
All together, differences in coefficients account for 13% of the gap.
You could also explain the 0.67 x 100, but it often makes little sense to interpret individual coefficients like that.
HTH

[Thea Fredriksen]

Thank you so much for explaining FernandoRios
As you seem to have great knowledge about the -oaxaca-, I would like to ask you another question.

I will perform another decomposition including a categorical variable for industry sector "indsect".
It is coded as followed:
(1) = Agriculture
(2) = Energy and Water
(3) = Manufacturing
and so on..

Can I include this in the decomposition like this:

bysort year: oaxaca lnwage age age2 indsect, by(female) noisily

Or should I create new industry dummies like this and then include those:

gen agriculture = 0
replace agriculture = 1 if indsect == 1

Thank you in advance!