Oaxaca Two Way Estimation: Setting Coefficients Equal Across Groups

[Herb Schuetze]

I am attempting to estimate a two-way oaxaca decomp in which the coefficents on the x-variables (except the intercept) are equal to the pooled coefficients across groups in calculating the unexplained portion. This is different than utilizing the pooled option, which utilizes the pooled coefficients to calculate the explained portion. Although, I will also use the pooled coefficients to calculate the explained part. In my case, only differences in the intercepts between the two groups would be included in the unexplained portion.

Group 1 (men): Y=a + Bp X
Group 2 (women) Y=A + Bp X

Oaxaca: (Xm - Xw)Bp + (a-A)

[George Ford]

Isn't that just a model with a group dummy?

Then use margins to compute the explained part using the means of each group (or the subpop option).

[George Ford]

Or maybe its the omega option you want.

[Herb Schuetze]

Thanks for your responses George. Correct it is just a model with a group dummy. I could code it up as you indicate but I was hoping for something easier that includes standard errors.

I will look closer at the omega option.

Herb

[George Ford]

The difference in pooled and omega is that the reference equation does not include the group dummy in the omega option. If you want all the coefficients but the intercept to be equal, then you want pooled. omega makes the intercepts the same.

Pooled does exactly what you want, I think. The only difference in the unexplained part is the intercept difference, since the other coefficients are identical and thus the other parts of the unexplained portion are 0.

I break down the calculations of Oaxaca decomposition here

https://www.statalist.org/forums/forum/general-stata-discussion/general/1739797-twofold-oaxaca-blinder-decomposition-interpretation

This might be of some use.

Run the model pooled and include split as an option. You'll see the first part of the decomposition (the difference in coefficients multiplied by the 1st group is 0).