128 Mohamad Fahmi
family background are unlikely to attend public schools, usually considered
the best schools in the country, than with private Islamic schools.
Based on positive and statistically significant on father education
background, students whose father that have upper secondary or higher
education background (FATHSHHE) have higher odds to attend public
schools than private schools. Moreover, students whose mother has upper
secondary or higher education background (MOTHSHHE) is most likely
to attend public school than private Islamic or private secular school.
Based on control variable for ability, students who have good early
academics ability or never failed in lower primary school have higher odds
to attend public schools than private schools. NEMSMP has a statistically
significant negative coefficient in all estimations as the higher score of
final exam means the higher odds of student attending public school. This
result is supported by the summary statistics by school type in Table 5.5
which indicates that students from private schools have lower scores on
national final exam (NEM) than the students from public schools. These
conditions occur as admission into most public schools require a minimum
grade on national final exam score.
Lower secondary school type coefficients are significant in terms of
the probability to choose upper secondary school. Students from private
secular and private Islamic schools are less likely to attend public schools
as compared to private schools. On the other hand, a student in private
Christian school prefers to attend public school than private Islamic school
even though he still has higher odds to attend private secular and private
Christian as compared to public school.
5.2 Higher Education Participation with Correction of
Selectivity Bias
As mentioned earlier, the selection of school particularly in public school
is not random and thus consequently there is a possibility of selection
bias in school choice. This chapter, following Lee (1983), will use the two
step technique to overcome the selection bias problem. The first step to
verify the selection is by creating selectivity variables. I use the results
from multinomial logit to calculate the selectivity variable, λij. Table 5.7
shows the estimations of probit model for higher education participation
or HEP which include λij as a regressor. The HEP probit estimation use
six exclusion restrictions in the structural equation: PRIFAIL, WORKSMP,