able indicating child(ren) under four years old living at home. The dummy Dit
takes the value one for AE participants after they have concluded their spell at
komvux and the variable ci is the number of credits attained divided by 500,
which is assumed equal to a year of AE.10 The vector zi contains controls for
the fractions of interrupted courses taken at compulsory level, as supplemen-
tary courses or as introductory courses. The Greek letters β, γ and λ are un-
known coefficients to be estimated and εit is the error term.
The coefficient γ will express the returns on annual earnings of a year of
full-time studies in AE. As Jacobson et al. (2005, p279) point out, the use of a
continuous variable of the credits makes the effects of AE identifiable also
from differences in the outcome of participants who obtain different number of
credits. In fact, given a proportional payoff to the amount of credits, it is not
even necessary to use data on non-participants to generate an estimate of the
yearly returns to AE, as the participants, in a manner of speaking, serve as their
own comparison group. However, the precision of the estimates improves when
one includes data on non-participants. A second strategy is to drop the continu-
ous variable ci and instead use binary variables, djit, which are one if the inter-
val number of credits is j, otherwise zero (Dit = ∑j djit ). The restriction of a
linear relationship between the number of credits and the returns is then
avoided, but at the cost of increased standard errors as the number of observa-
tions is reduced for each estimated γ coefficient.
Table 6 presents coefficient estimates pertaining to a year of AE credits.
The parameters of the first four columns are, at least conceptually, straight-
forward to interpret as they reasonably imply an addition to the individual’s
level of formal education. These estimations imply positive effects that vary
from 5 to 15 per cent. Credits earned by individuals with a prior three-year up-
per secondary schooling (the last two columns) are more complicated to inter-
pret as they may reflect an improvement of earlier grades and/or a re-direction
of an earlier education, unless the aim was to complete the optional fourth year
10 The value of the variable ci is zero both for individuals never registered in AE and for those
who were registered but never gained any credits.
IFAU - Does adult education at upper secondary level influence annual wage earnings?
25