5.2 Individuals with no enrolment in higher
education
The analysis in this section is restricted to those that at some point were regis-
tered in higher education. The sample concerned is the one described in section
4.2. The main purpose is to explore whether individuals with and without prior
AE differ in their study achievements at university and/or in their wage earn-
ings premium of higher education.12 The latter may occur if a positive effect of
education on earnings decreases with age or if AE individuals tend to engage in
education due to dismal employment prospects. A similar implication would
arise if AE is of insufficient quality as preparation for university studies.
To control for ability when comparing the different outcomes, the grade
point average (henceforth GPA) from upper secondary school is used. The
GPA variable is positively correlated with the number of years of higher stud-
ies. When there is no support in GPA values across groups with different years
of schooling, a regression estimator extrapolates over regions of no support to
achieve comparability between the groups. If one adjusts the sample to retain
common support, Rubin (1973a, 1973b) showed that bias is reduced as well as
the sensitivity of the estimator with regard to incorrect functional form as-
sumptions. For this reason, the samples in the following are constructed so that
the GPA values overlap between the groups with different number of years of
higher studies. Individuals are excluded if their GPA is lower than the 5th per-
centile in the group classified with four years of university studies or if their
GPA is higher than the 95th percentile of the group classified with less than one
year of university studies. The limit values of the GPA are calculated sepa-
rately for each sample but in all cases stay within the range from 2.80 to 4.19,
excluding roughly one fourth of the observations.13 Descriptive characteristics
of restricted samples are presented in Table A.3 in the Appendix.
12 Regression-like estimations of the probability of continuing from AE to higher education are
not very interesting as the AE in many cases was a precondition for the individual to become eli-
gible for higher studies.
13 The location of the studies is not controlled for. Lindahl and Regnér (2006) find the payoff to
studies to differ between different locations but Eliasson (2006), when using the GPA as a con-
trol for otherwise unobserved characteristics, finds the effects associated with location to disap-
pear.
IFAU - Does adult education at upper secondary level influence annual wage earnings?
33