minimal level of the grading when youth education is evaluated. Another ex-
planation may be that wage dispersion has grown in Sweden since the eco-
nomic recession at the start of the 1990’s (e.g. Lundborg, 2005), and the studies
referred to in section 3 are based on data collected before this occurred. A more
problematic aspect is that the dynamics of wage earnings may be different in
different parts of the wage earnings distribution. As AE individuals more often
are in the lower half, it is not unreasonable to think that their average employ-
ment level is influenced more positively by the overall economic recovery from
1996 and onwards. The fixed effects regression would not be able to properly
capture such a difference and estimates would partly reflect time-variant ef-
fects. In the literature evaluating Labour Market Training such pro-cyclical re-
sults seem to have been present (see Andrén and Gustafsson, 2005, and discus-
sion in Stenberg, 2007). If the positive effects in Table 6 are driven mainly by
low earners, applying absolute values of the earnings variable should attenuate
the estimates. This happens for the full samples with a two-year upper secon-
dary level. The coefficient becomes significantly negative for males (-2,854)
and relatively modest for females (1,773). These results change back to become
similar to those in Table 6 when individuals with no annual earnings above
SEK 100,000, or individuals with UBS, are excluded. Using absolute values of
earnings for those with a three-year upper secondary level, the point estimates
of the full sample become significantly negative, -22,776 for males and -4,525
for females, and remain roughly unchanged, although for females insignificant,
when the sampling conditions are altered.
Table 7 presents coefficient estimates of the version with binary variables
for intervals of accomplished credits. The intervals are the same as those in the
descriptive section (see Table 3), except that the two groups with more than
1,000 credits are merged into one group. To ease the exposition, the mostly in-
significant results of the sample with a prior three-year upper secondary educa-
tion are not displayed.
For individuals with a prior compulsory level, the coefficient values for
males and females still reflect substantial payoffs but also rather large standard
errors (there are only 36 individuals in the group with more than 1,000 credits).
There are no significantly positive returns to less than one whole semester of
AE. This is along the lines of what was reported by Stenberg and Westerlund
(2007) for long-term unemployed persons in the AEI as well as in the survey of
studies on community college effects by Grubb (2002).
28
IFAU - Does adult education at upper secondary level influence annual wage earnings?