I'm currently estimating some earning equations and controlling for unexplained earnings for public employees at State level here in Brazil. One procedure common to this approach that I found on the most recent papers about the subject is to correct the earning equations with a biprobit model, controlling for the probability of the individual to be employed and the probability of being a public employee as well.
These are the results that I have for my whole sample for the year 2015:
Code:
Survey: Bivariate probit regression Number of strata = 728 Number of obs = 2,190,952 Number of PSUs = 16,305 Population size = 1,069,853,667 Subpop. no. obs = 180,108 Subpop. size = 103,535,992 Design df = 15,577 F( 28, 15550) = 1091.28 Prob > F = 0.0000 ---------------------------------------------------------------------------------- | Linearized | Coef. Std. Err. t P>|t| [95% Conf. Interval] -----------------+---------------------------------------------------------------- emprego | nfilhos2 | -.2114478 .0128748 -16.42 0.000 -.2366839 -.1862116 nfilhos3_5 | -.1098781 .0126787 -8.67 0.000 -.1347299 -.0850264 nfilhos6_12 | -.0974262 .0080992 -12.03 0.000 -.1133016 -.0815508 nfilhos13_17h | -.0164063 .0113128 -1.45 0.147 -.0385808 .0057682 nfilhos13_17m | -.0086634 .0127159 -0.68 0.496 -.033588 .0162611 chefe | .3851882 .0106939 36.02 0.000 .3642269 .4061496 negro | -.0563385 .0086413 -6.52 0.000 -.0732764 -.0394005 renda_extra | -.0001249 5.15e-06 -24.27 0.000 -.000135 -.0001148 estudo | .1492859 .0037893 39.40 0.000 .1418584 .1567133 idade | .1454735 .0019834 73.34 0.000 .1415857 .1493613 | c.estudo#c.idade | -.0006611 .0000822 -8.04 0.000 -.0008222 -.0005 | c.idade#c.idade | -.0018171 .0000225 -80.87 0.000 -.0018611 -.0017731 | estudante | -.3296371 .0129974 -25.36 0.000 -.3551135 -.3041607 membros | .0038305 .0034291 1.12 0.264 -.002891 .010552 _cons | -3.929445 .0538253 -73.00 0.000 -4.034949 -3.823942 -----------------+---------------------------------------------------------------- pub_est | nfilhos2 | -.1002891 .02476 -4.05 0.000 -.1488216 -.0517565 nfilhos3_5 | -.0979429 .0255955 -3.83 0.000 -.1481131 -.0477726 nfilhos6_12 | -.0527653 .0162168 -3.25 0.001 -.0845522 -.0209784 nfilhos13_17h | -.0236773 .0223264 -1.06 0.289 -.0674398 .0200851 nfilhos13_17m | -.0065224 .0235071 -0.28 0.781 -.0525991 .0395543 chefe | .1427142 .0150768 9.47 0.000 .1131619 .1722665 negro | .05077 .0159538 3.18 0.001 .0194987 .0820413 renda_extra | -.0000702 3.92e-06 -17.91 0.000 -.0000779 -.0000625 estudo | .2175808 .0101962 21.34 0.000 .1975952 .2375665 idade | .1371285 .0045807 29.94 0.000 .1281497 .1461072 | c.estudo#c.idade | -.0012952 .0001982 -6.53 0.000 -.0016837 -.0009067 | c.idade#c.idade | -.0012525 .0000367 -34.09 0.000 -.0013245 -.0011805 | estudante | .1469598 .0229742 6.40 0.000 .1019278 .1919918 membros | .0251703 .006382 3.94 0.000 .0126607 .0376799 _cons | -7.12937 .1658712 -42.98 0.000 -7.454497 -6.804243 -----------------+---------------------------------------------------------------- /athrho | 1.386635 .0257015 53.95 0.000 1.336257 1.437013 -----------------+---------------------------------------------------------------- rho | .8824283 .0056883 .8707703 .8930948 ----------------------------------------------------------------------------------
The model seems ok and now I need to obtain the Inverse Mills Ratio to plug in my earning equations to correct for selection bias. Seems like the -biprobit- command does not have a built in option to obtain that, so, any thoughts on how I cant obtain it and plug in my other equation?
Thanks,
Fernando Martins.
0 Response to Sample selection correction after biprobit
Post a Comment