Hi,
I am studying the wage differentials between graduates working in the public and private sectors using Oaxaca-Blinder decomposition method. Explanatory variables includes academic ability, English competency, socio-economic variables, and job characteristics. I am using the oaxaca package with twofold decomposition technique (Jann, 2008).
The output is given as follows:
Looking at the output, the difference between the mean log(wage) for public minus private sector is not significant. However, the explained and the unexplained parts are significant and constitute nearly the same bearing. Can I say that the difference, albeit insignificant, is accounted for by equal amount of characteristics effects (the explained portion) and some unexplained components?
Take for example the variable Male which contribute both to the explained (-0.0130***) and unexplained parts (-0.0499***). What can be said about the contribution of Male dummy towards the difference in the earnings between the two sectors? The negative value of Male coefficient in the explained part may suggest that the difference in the earnings may be due to being a female. But what can be said about the Male coefficient in the unexplained part?
Also, academic performance GPA in the unexplained part is negative. Is it safe to assume that the wage difference is due to some unexplained attributes on account of lower GPA?
Thank you.
Jann, B. (2008). The Blinder-Oaxaca decomposition for linear regression models. The Stata Journal, 8(4), 453-479.
I am studying the wage differentials between graduates working in the public and private sectors using Oaxaca-Blinder decomposition method. Explanatory variables includes academic ability, English competency, socio-economic variables, and job characteristics. I am using the oaxaca package with twofold decomposition technique (Jann, 2008).
The output is given as follows:
Code:
global x4 = “gpa denglish* dcourses* male age chinese dfaminc* urban fulltime permanent professional”
oaxaca gradinc $x4, by(public) weight(0) relax detail(English:denglish*, Faminc=dfaminc*, Courses=dcourses*)
-----------------------------------------
(1)
gradinc
-----------------------------------------
overall
group_1 3.969*** (0.0323)
group_2 4.009*** (0.00993)
difference -0.0399 (0.0338)
explained 0.211*** (0.0378)
unexplained -0.251*** (0.0301)
-----------------------------------------
explained
gpa 0.0248 (0.0142)
English 0.00372 (0.00432)
Courses 0.441*** (0.0289)
male -0.0130*** (0.00300)
age 0.0229*** (0.00402)
chinese -0.0162* (0.00751)
Faminc 0.000470 (0.00302)
urban -0.00764** (0.00262)
fulltime -0.134*** (0.00893)
permanent -0.183*** (0.0117)
professional 0.0722*** (0.00688)
-----------------------------------------
unexplained
gpa -1.140*** (0.194)
English -0.0297 (0.0687)
Courses -0.919*** (0.235)
male -0.0499*** (0.0146)
age 2.025*** (0.445)
chinese -0.0898*** (0.0191)
Faminc 0.00489 (0.0653)
urban -0.0659*** (0.0148)
fulltime -0.274*** (0.0377)
permanent 0.209*** (0.0242)
professional -0.00568 (0.0272)
_cons 0.0317 (0.570)
-----------------------------------------
N 28570
-----------------------------------------
Standard errors in parentheses
* p<0.05, ** p<0.01, *** p<0.001
Take for example the variable Male which contribute both to the explained (-0.0130***) and unexplained parts (-0.0499***). What can be said about the contribution of Male dummy towards the difference in the earnings between the two sectors? The negative value of Male coefficient in the explained part may suggest that the difference in the earnings may be due to being a female. But what can be said about the Male coefficient in the unexplained part?
Also, academic performance GPA in the unexplained part is negative. Is it safe to assume that the wage difference is due to some unexplained attributes on account of lower GPA?
Thank you.
Jann, B. (2008). The Blinder-Oaxaca decomposition for linear regression models. The Stata Journal, 8(4), 453-479.
Comment