Indonesian Higher Education 119
4. DATA
4.1 Description of Sample and Sampling Procedure
To estimate the probability of higher education participation in public and
private school groups, I use data from the last three waves of the IFLS. The
IFLS is a continuing longitudinal survey that provides rich information
collected at the individual, household, and community level and contains
a variety of economic and non-economic indicators. The IFLS is based on
a stratified sample of provinces with random sampling of households
within provinces. The stratified sampling of provinces was used not only
to capture the cultural diversity of Indonesia but also to be cost effective.^2
The survey was conducted in thirteen provinces with the sample being
representative of about 83 per cent of the Indonesian population. This
chapter exploits waves 2–4 of the IFLS, omitting wave 1 to data differences
between the later 3 waves and the first wave.
I follow the sample selection of Flanerry (2009) focusing on a sample
of respondents whose age are from 17 to 24 years old, have finished
upper secondary education and face the decision to participate in higher
education. Adults over 24 years old are excluded since the decision for
these respondents to participate in higher education may be influenced by
a greater working opportunity and some other factors that are incorporated
in the model. Some respondents in the 17–24 years age cohort were either
still studying or did not finish upper secondary school and thus are also
excluded from the sample.
Since the IFLS is a longitudinal survey, there may be multiple
observations of individual respondent’s probabilities that a respondent
appears on more than one wave. I provide a particular treatment to classify
those respondents. A respondent who was interviewed in all three waves is
classified based on the wave which his or her age is within the 17–24 years
old bracket. For instance, a respondent whose age is 9 years old in IFLS2
and 12 and 19 years old in IFLS3 and IFLS4 would be classified as being
in the IFLS4 sample. If a respondent appears in two or three waves and
his or her age is within the age cohort in two waves, then I focus on the
latest wave on the assumption that the latest wave is the most up-to-date.
Table 5.1 presents the number of respondents by IFLS wave.
The full sample consists of 88,030 observations from three waves.
I reduced the sample to 1,400 observations by making the exclusions that