Dear all,
I use the oaxaca command as presented in Jann, Ben (2008). The Blinder-Oaxaca decomposition for linear regression models. The Stata Journal 8(4): 453-479. I am not sure if I use heckman option correctly, since I get quite strange results. Here is the problem in more detail.
OLS regression gives me adjusted gender pay gap (GPG) of 14% and stat. sig. Heckman selection model gives me GPG of 5.4%, i.e. GPG is lower when I take into account selection in the labour market. Oaxaca command gives me the same adjusted GPG as OLS as expected.
The problem is when I use oaxaca with heckman option. First of all, I am not sure should I use only one moedel (male or female) for selection or both models?
When I use males model for selection the GPG with selection increases to 32%, female selection model gives me results of 17%, and when both models are used I get 0.8% and not significant.
Should oaxaca with heckman give approximately the same coefficient as the female cofficient in Heckman? Taking into account other previous research on this topic for my country, GPG with selection should be lower than without selection. Anyway, the result of 32% and of 0.8% are quite unrealistic.
I use STATA15, if that maybe metter.
Thank you in advance for your answer.
Best,
Aleksandra
I use the oaxaca command as presented in Jann, Ben (2008). The Blinder-Oaxaca decomposition for linear regression models. The Stata Journal 8(4): 453-479. I am not sure if I use heckman option correctly, since I get quite strange results. Here is the problem in more detail.
OLS regression gives me adjusted gender pay gap (GPG) of 14% and stat. sig. Heckman selection model gives me GPG of 5.4%, i.e. GPG is lower when I take into account selection in the labour market. Oaxaca command gives me the same adjusted GPG as OLS as expected.
The problem is when I use oaxaca with heckman option. First of all, I am not sure should I use only one moedel (male or female) for selection or both models?
When I use males model for selection the GPG with selection increases to 32%, female selection model gives me results of 17%, and when both models are used I get 0.8% and not significant.
Should oaxaca with heckman give approximately the same coefficient as the female cofficient in Heckman? Taking into account other previous research on this topic for my country, GPG with selection should be lower than without selection. Anyway, the result of 32% and of 0.8% are quite unrealistic.
I use STATA15, if that maybe metter.
Thank you in advance for your answer.
Best,
Aleksandra
