Blinder-Oaxaca Decomposition interpretation

[Altaf Alam]

I am using Blinder Oaxaca decomposition to estimate urban-migrant and urban wage gap. In a two fold decomposition, I am getting negative value of Unexplained part. Explained part is positive and both (explained and unexplained) are significant. What is the interpretation of negative unexplained if the value is in antilog form as the main dependent variable was log wage. Control variables are education, experience etc.. .

[Dung Le]

Hi Altaf,

You may want to read these papers by Ben Jann: https://boris.unibe.ch/117107/1/2-oaxaca.pdf and https://journals.sagepub.com/doi/10....867X0800800401

In addition, you also can post Stata commands that you used to evaluate your data and details of your results so people here are able to help you out.

DL

[Altaf Alam]

Getting negative value of the unexplained part of Oaxaca decomposition. Explained part has the value more than the log wage difference of migrant and urban worker. Stata log file attached.

[Altaf Alam]

Following is my code to do Blinder Oaxaca Decomposition
[xi: oaxaca lhrw uni yearofschool i.occup i.industry i.jobtype tenure maris exper expr2, by(migrant) weight(1) detail
oaxaca, eform]
However, I am getting negative value of the unexplained part. The reason of getting negative value, is the constant term. My two groups are migrant and urban work force. Migrant group has higher constant term than urban group. In the unexplained part constant term is the difference between the urban group constant term minus migrant group constant term. This negative value is so high that it is making the whole unexplained part negative. Is there any solution?
Example: After Oaxaca decomposition
log Wage Difference 0.20
Explained part 0.30
Unexplained Part - 0.10
Value of the constant term of Urban group regression 1.29
Value of the constant of Migrant group regression 1.87
Can any statalist member help me?

[Weiwen Ng]

Following is my code to do Blinder Oaxaca Decomposition
[xi: oaxaca lhrw uni yearofschool i.occup i.industry i.jobtype tenure maris exper expr2, by(migrant) weight(1) detail
oaxaca, eform]
However, I am getting negative value of the unexplained part. The reason of getting negative value, is the constant term. My two groups are migrant and urban work force. Migrant group has higher constant term than urban group. In the unexplained part constant term is the difference between the urban group constant term minus migrant group constant term. This negative value is so high that it is making the whole unexplained part negative. Is there any solution?
Example: After Oaxaca decomposition
log Wage Difference 0.20
Explained part 0.30
Unexplained Part - 0.10
Value of the constant term of Urban group regression 1.29
Value of the constant of Migrant group regression 1.87
Can any statalist member help me?

This means that the control variables explain more than the whole difference in (log) wages. In other words, if you gave migrant workers the same values of the control variables as urban workers, you'd expect migrant workers to have greater wages than urban workers.

Difference = Explained + Unexplained

Normally, explained is a fraction of the total difference. Here, that fraction is actually over 1.

If there is a data error, that could explain this. If there isn't a data error, then there's nothing to solve.

[Altaf Alam]

Hi I want.to know the STATA command for doing Cotton (1988) type Oaxaca decomposition mentioned in Jann Paper. Should I calculate the weight separately considering the sample sizes of each group and multiply by coefficients or any easy way..