Dear all,
I know I am asking a lot with the following post, but I would be very thankful for any help: For my bachelor thesis, I am trying to estimate a wage offer function based on a Panel Data Set. I came across an estimator discussed in Wooldridge (1995) that corrects for sample selection bias in panel data models. Applying that estimator itself did not cause me any troubles. However, since it is a two-stage model, a consistent estimation of the standard errors of the second stage would need to take into account the estimation-nature of the first-stage results included in the second-stage results.
In a previous post (http://www.statalist.org/forums/foru...eckman-process), a user was struggling with bootstraping consistent estimators. However, Wooldridge (1995) proposed an own estimator for those standard errors, that I fail to apply in Stata.
The idea of the Wooldridge (1995) estimator is to (for each t) obtain Tobit estimates of the selection variable (in my case, the hours worked), and then use the difference between the selection variable and the fitted xb values of the Tobit estimation in the second stage (wage offer) estimation.
The procedure proposed by Wooldridge (1995) to obtain consistent standard errors is the following (step (iii) is the actual estimator):
This is what i did so far for steps (i) and (ii) (stata Version: 13.0):
But I really have no idea how to continue. Can anyone first help me to obtain the z_it and then the estimator of D?
Thank you so much in advance!
Johann
Wooldridge, J. M. (1995), ‘‘Selection Corrections for Panel Data Models under Conditional Mean Independence Assumptions,’’ Journal of Econometrics 68, 115-132.
I know I am asking a lot with the following post, but I would be very thankful for any help: For my bachelor thesis, I am trying to estimate a wage offer function based on a Panel Data Set. I came across an estimator discussed in Wooldridge (1995) that corrects for sample selection bias in panel data models. Applying that estimator itself did not cause me any troubles. However, since it is a two-stage model, a consistent estimation of the standard errors of the second stage would need to take into account the estimation-nature of the first-stage results included in the second-stage results.
In a previous post (http://www.statalist.org/forums/foru...eckman-process), a user was struggling with bootstraping consistent estimators. However, Wooldridge (1995) proposed an own estimator for those standard errors, that I fail to apply in Stata.
The idea of the Wooldridge (1995) estimator is to (for each t) obtain Tobit estimates of the selection variable (in my case, the hours worked), and then use the difference between the selection variable and the fitted xb values of the Tobit estimation in the second stage (wage offer) estimation.
The procedure proposed by Wooldridge (1995) to obtain consistent standard errors is the following (step (iii) is the actual estimator):
This is what i did so far for steps (i) and (ii) (stata Version: 13.0):
Code:
forvalues i = 1991 1992 to 2012 {
quietly tobit hours x1-x5 if JAHR==`i', ll(0)
quietly predict xb_hat_`i', xb
quietly gen delta_hat_d_`i' = hours - xb_hat_`i'
quietly replace delta_hat_d_`i' = 0 if JAHR!=`i'
quietly drop xb_hat_`i'
}
reg wage z1-z4 delta_hat_d_1991-2012 if wage>0
Thank you so much in advance!
Johann
Wooldridge, J. M. (1995), ‘‘Selection Corrections for Panel Data Models under Conditional Mean Independence Assumptions,’’ Journal of Econometrics 68, 115-132.
