Hi. I am examining the impact of the migration of one member on children in the migrant's household. I am using the following regression to measure the impact of migration on school dropouts:

Using 2SLS
Code:
#delimit ; 
ivregress 2sls dropout (migrant = mig_percent nb_children nb_ymale nb_yfemale nb_elderly avg_school) age age2 eldest extended fth_absent i.fteducst i.mteducst  [pweight=expan_indiv], vce (cluster hhid) first;
#delimit cr
and using IV probit:
#delimit ;
ivprobit dropout (migrant = mig_percent nb_children nb_ymale nb_yfemale nb_elderly avg_school) age age2 eldest extended fth_absent i.fteducst i.mteducst [pweight=expan_indiv], vce (cluster hhid) first;
#delimit cr
Here is a description of the variables:
dropout is a binary variable = 1 if the child has dropped out of school
migrant is a binary variable = 1 if the child lives in a household with a migrant
nb_children, nb_ymale, nb_yfemale, nb_elderly represent the composition of the household
eldest is a binary variable = 1 if the child is the eldest child
extended is a binary variable = 1 if the child lives in an extended household
fteducst and mteducst are father's and mother's educational levels respectively



My problem is that when using the same equation for 2SLS and IVprobit, I get less observations in the IVprobit regression as you can see from the results

2SLS IV probit
migrant -0.216*** -1.530
(0.07) (1.11)
age -0.037*** 0.253*
(0.00) (0.13)
Age2 0.002*** -0.001
(0.00) (0.00)
eldest -0.005 -0.047
(0.00) (0.08)
extended -0.009 0.012
(0.01) (0.12)
Fth_absent 0.074*** 0.523*
(0.02) (0.31)
fteducst
Reads & Writes -0.022* -0.194*
(0.01) (0.11)
Less than Intermediate -0.044*** -0.448***
(0.01) (0.12)
Intermediate -0.046*** -0.669***
(0.01) (0.12)
Above Intermediate -0.062*** -1.368***
(0.01) (0.36)
University -0.053*** -1.013***
(0.01) (0.28)
Post-Graduate -0.052*** 0.000
(0.01) (.)
mteducst
Reads & Writes -0.006 0.024
(0.01) (0.14)
Less than Intermediate -0.020** -0.200
(0.01) (0.13)
Intermediate -0.029*** -0.702***
(0.01) (0.14)
Above Intermediate -0.025*** -0.435
(0.01) (0.44)
University -0.021*** 0.000
(0.01) (.)
Post-Graduate -0.011 0.000
(0.01) (.)
Constant 0.200*** -4.979***
(0.03) (0.92)
Observations 12407 10816

I noticed that "Post-Graduate" in father's education (fteducst) is empty in the ivprobit regression. I also noticed that "University" and "Post-Graduate" in mother's education (mteducst) are empty in the ivprobit regression.

Stata also showed the following comment before the IVprobit results
Code:
note: 7.fteducst != 0 predicts failure perfectly
      7.fteducst dropped and 94 obs not used

note: 6.mteducst != 0 predicts failure perfectly
      6.mteducst dropped and 1540 obs not used

note: 7.mteducst != 0 predicts failure perfectly
      7.mteducst dropped and 82 obs not used
I was wondering if anyone could help me get the same number of observations in both regressions. Thanks in advance