Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Double Selection - Mill's ratios computation

From   Fabrizio Colella - 1348010 <>
Subject   st: Double Selection - Mill's ratios computation
Date   Thu, 12 Dec 2013 20:56:39 +0100 (CET)

Dear all,

I am performing an Oaxaca Blinder wage decomposition among women working part-time and women working full-time, 
with a double switching model as selection-correction. 

To do that I performed first two -probit- regressions to "correct" for two types of selection:
1) participation vs. non-participation in the labor market
2) full-time vs. part-time – (CONDITIONAL ON PARTICIPATION)
I calculated the two mills ratios and I used them as covariates in the –oaxaca-.

As you know there are many different methods to control for selection using the inverse Mill's ratios (Heckman, Greene, Maddala),
Below there is the code I am using, anyway I am not sure it works. Please, could anybody check it?

*** Selection  
probit labor_force_partecipation age_25_35 … … health_prob, robust iterate(10)
predict Xb1 , xb
gen mills1 = .
replace mills1 = normalden(-Xb1)/(1-(normal(-Xb1))) if lfp==1 //workers
probit part_time age_25_35 … … health_prob, robust iterate(10)
predict Xb2 , xb
gen mills2 = .
replace mills2 = normalden(-Xb2)/(1-(normal(-Xb2))) if part_full_only==1 //part_timers
replace mills2 = -normalden(-Xb2)/((normal(-Xb2))) if part_full_only==0 //full_timers
*** OB decomposition ***
oaxaca ln_hourly_wage ed_primary … … worker mills1 mills2 [pweight=pb040], by(part_full_only) weight(0)

Suggestions will be highly appreciated.

Thank you in advance

Fabrizio Colella
Bocconi University and
Fondazione Rodolfo DeBenedetti
Milan, Italy

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index