代码拉取完成，页面将自动刷新
Kamila Kolpashnikova
May 11, 2016
The data here is derived from the American Time Use Survey. For the coding procedures, contact me via email.
You need to download the ATUS_full3.dta file from my github to your working directory.
use ATUS_full3.dta, clear
What is the Oaxaca-Blinder decomposition method? The Blinder-Oaxaca pooled method includes the differentiation between the explained part of the gap and the unexplained part of the gap. The definitions for explained and unexplained decomposition can be represented in the following way:
$R = [E(X_1) - E(X_2)] '\beta^* + [E(X_1) ' (\beta_1 - \beta^) + E(X_2) ' (\beta^ -\beta_2)]$, where
$[E(X_1) - E(X_2)] '\beta^$ is the explained part and $[E(X_1) ' (\beta_1 - \beta^) + E(X_2) ' (\beta^* -\beta_2)]$ is unexplained.
Where $\beta^*$ is non-discriminatory coefficients vector (Oaxaca and Ransom 1994; Jann 2008).
For a variation of the pooled decomposition: $\beta^* = W\beta_A + (I-W)\beta_B$; where $W$ – relative weights given to coefficients of group A (Oaxaca and Ransom 1994). $\hat{W}=\Omega=(X_A’X_A+X_B’X_B)^{-1}X_A’X_A$. In this tutorial, I used the pooled method of constructing the non-discriminatory wage structures as proposed by Oaxaca and Ransom (1994). The Fig . below represents such a situation where $W=0.5$. For more detail, please see Elder, Goddeeris, and Haider (2010).
To see how much of the gender gap the difference in the men’s and women’s mean on a particular variable produces in an outcome variable, we multiply the mean difference by the slope from the hypothesized non-discriminatory slope $ß^*$, since the slope tells us the rate of return to a unit change in the variable (see Fig.). This produces the amount of the dependent variable which we can explain with the difference in the independent variable. The unexplained part therefore, is the difference from the explained part to the highest point in the interval of the slope above and the difference to the lowest point in the interval from the slope below. This part is usually attributed to the effects of discrimination and unobserved variables (Jann 2008).
There are a few variations of the (two-fold) Oaxaca-Blinder decomposition:
To use Group 1 coefficients as reference, you need the Stata command weight(1).
To use Group 2 coefficients as reference, you need the Stata command weight(0).
If you need the coefficients from a pooled regression as reference as specified above (Neumark 1988) and excluding the group indicator, use the Stata command omega.
If you need the coefficients from a pooled regression as reference as specified above (Jann 2008) and including the group indicator, use the Stata command pooled.
You can use a variation of the following command to run a pooled two-fold Oaxaca-decomposition:
#delim ;
oaxaca lnDVCOOK IncomeTransfer Weekday Year
[pw=Weight] if Married==1
& HhldSize>1 & BornInUSA==1, by(Female) pooled robust;
Blinder-Oaxaca decomposition Number of obs = 39595
Model = linear
Group 1: Female = 0 N of obs 1 = 13651
Group 2: Female = 1 N of obs 2 = 25944
--------------------------------------------------------------------------------
lnDVCOOK | Coef. Std. Err. z P>|z| [95% Conf. Interval]
---------------+----------------------------------------------------------------
overall |
group_1 | 3.396838 .0111317 305.15 0.000 3.37502 3.418656
group_2 | 3.874199 .0075316 514.39 0.000 3.859438 3.888961
difference | -.4773614 .0134402 -35.52 0.000 -.5037037 -.4510191
explained | -.0646089 .003788 -17.06 0.000 -.0720333 -.0571844
unexplained | -.4127525 .0137204 -30.08 0.000 -.439644 -.3858611
---------------+----------------------------------------------------------------
explained |
IncomeTransfer | -.0670399 .0036881 -18.18 0.000 -.0742684 -.0598115
Weekday | .0016783 .0007919 2.12 0.034 .0001262 .0032305
Year | .0007527 .0003751 2.01 0.045 .0000176 .0014879
---------------+----------------------------------------------------------------
unexplained |
IncomeTransfer | -.0288756 .0050905 -5.67 0.000 -.0388528 -.0188985
Weekday | -.1026541 .0180548 -5.69 0.000 -.1380407 -.0672674
Year | 22.53606 7.198423 3.13 0.002 8.427406 36.64471
_cons | -22.81728 7.197366 -3.17 0.002 -36.92386 -8.710701
--------------------------------------------------------------------------------
You can also use detail specification to see the effect of the combined variables.
#delim ;
oaxaca lnDVCOOK lnDVPAID Children HhldSize Under5
FullTime PartTime Other Weekday Year[pw=Weight] if Married==1
& HhldSize>1 & BornInUSA==1, by(Female) pooled robust relax
detail(Emp: FullTime PartTime Other);
Blinder-Oaxaca decomposition Number of obs = 16041
Model = linear
Group 1: Female = 0 N of obs 1 = 6727
Group 2: Female = 1 N of obs 2 = 9314
------------------------------------------------------------------------------
lnDVCOOK | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
overall |
group_1 | 3.178306 .0149955 211.95 0.000 3.148915 3.207696
group_2 | 3.607364 .0113488 317.86 0.000 3.585121 3.629608
difference | -.4290586 .0188058 -22.82 0.000 -.4659173 -.3921998
explained | -.0310219 .005491 -5.65 0.000 -.0417841 -.0202598
unexplained | -.3980366 .0191087 -20.83 0.000 -.4354891 -.3605842
-------------+----------------------------------------------------------------
explained |
lnDVPAID | -.020901 .0030364 -6.88 0.000 -.0268522 -.0149498
Children | .0052965 .0019284 2.75 0.006 .0015169 .0090761
HhldSize | .0029706 .0015543 1.91 0.056 -.0000757 .0060169
Under5 | -.0013118 .0013757 -0.95 0.340 -.0040082 .0013846
Emp | -.0188331 .0042856 -4.39 0.000 -.0272327 -.0104335
Weekday | .0006296 .0005001 1.26 0.208 -.0003506 .0016098
Year | .0011273 .0006287 1.79 0.073 -.000105 .0023595
-------------+----------------------------------------------------------------
unexplained |
lnDVPAID | -.0109932 .1405219 -0.08 0.938 -.2864109 .2644246
Children | .0283648 .030039 0.94 0.345 -.0305105 .0872402
HhldSize | -.1394688 .0712782 -1.96 0.050 -.2791716 .0002339
Under5 | .0127992 .009905 1.29 0.196 -.0066142 .0322126
Emp | .0507413 .0529143 0.96 0.338 -.0529688 .1544514
Weekday | -.108484 .0446155 -2.43 0.015 -.1959288 -.0210392
Year | 27.72879 10.29912 2.69 0.007 7.542895 47.91469
_cons | -27.95979 10.29938 -2.71 0.007 -48.14619 -7.773381
------------------------------------------------------------------------------
Emp: FullTime PartTime Other
此处可能存在不合适展示的内容，页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容，可点击提交进行申诉，我们将尽快为您处理。