due to macro-economic fluctuations. Also, it is an open question whether age
per se and/or prior work experience plays a part in the payoff to AE.
Selection bias originating from unobservable characteristics typically causes
particular concern. One way to address this problem is to use difference-in-dif-
ferences regression estimators. The pre-AE annual earnings of the individual
are then used to take account of time-invariant unobservable variables that in-
fluence annual earnings post-AE. A special case of such a regression estimator
is the individual fixed effects model described in section 5.1. In section 5.2, the
focus is on individuals participating in higher education. Annual earnings be-
fore education are then more problematic to use as many individuals go straight
from upper secondary school to university studies. The strategy is instead to
make estimates on the cross section of wage earnings registered in 2002 and to
use grade point averages from upper secondary school as a control for ability
bias.
5.1 Individuals with no enrolment in higher
education
This part deals with estimations of how upper secondary credits attained at
komvux influence registered annual wage earnings from 1990 until 2001.
Throughout, results are presented for samples that separate females and males
and by prior educational level, defined as that reported in SUN1990 (Table 1).
OLS regression estimates of individual fixed effects models are used to control
for time-invariant unobserved heterogeneity across individuals. Included are
also yearly dummies to control for economic fluctuations. The fixed effects
model has the advantage that it also allows for controls indicating temporary
variations in the individual’s family situation. Formally, the model can be
written
yi = α+ μi + θ + βbi 1 + γDi ci + λDi zi + εi
it i t it -1 it i it i it
where yit represents the logarithmic annual wage earnings of individual i at time
t. Explanatory variables are the individual specific fixed effects μi, the time-
specific effects θt and the individual characteristic bit-1 which is a binary vari-
24
IFAU - Does adult education at upper secondary level influence annual wage earnings?