130 Mohamad Fahmi
The coefficients on λij are statistically significant for public, private
secular, and private Islamic schools. These evidence imply that there
are selection biases in upper secondary school choice. Coefficients on
selectivity variable for public schools and private Christian schools are
negative which indicate that unobservable characteristics of students in
these school types tend to have lower probability to participate in higher
education. On the other hand, the negative coefficient on λij in private
secular and private Islamic schools imply that unobservable characteristics
of students in these schools tend to have higher probability to participate
in higher education.
Prior to discussing the results in probit model with selectivity
correction, I compare the results to the probit model without selectivity
bias. The results of probit model without selectivity bias correction are
presented in Table 5.8. Whether or not the selectivity bias coefficients are
significant, the insertion of these variables tends to alter other coefficient
estimates. Generally, coefficients in all estimations decreased. Some of the
estimations have more significant variables. One that changed dramatically
is the private Christian group; the probit estimation without the selection
bias treatment has only three significant variables, but, however, after the
addition by selection variable the significant variables increased to nine.^3
In contrast, in the private secular group the number of significant variables
decreases from 10 to 9. One variable that changed from significant at 1 per
cent level to insignificant is SMAFAIL.
The results of probit estimation with selectivity correction provide
important findings about the determinants of higher education participation
of students from each type of school. Some of the determinants have
the same effect as the decision of student from different type of school
whereas some others do not. The variables that have same effect to the
student across different type of school are earnings forgone (FOREGONE),
parents’ upper secondary and higher education background (FATHSHHE
and MOTHSHE), number of sibling (SIBLING), sex (MALE), and use of
language (LANGINDO).
The coefficients on earning foregone (FOREGONE), as expected, are
negative and significant in all type of schools which means that students
have disincentives to attend higher education from positive change on
earnings forgone. Students from private Christian school are the most
sensitive to the change in earnings foregone as its coefficient on FORGONE
is the lowest or –0.419.