Demographic Research: Volume 21, Article 6
give estimates ranging between 1.80 and 1.985. The remaining education categories
generally encompass only a small number of events6. The results for these categories
probably should not be over-interpreted. Even for these though, deviations appear to be
quite small. So, the question is why the estimates derived from the imputed histories are
so close to the ones given by the original histories (with the exception of the effect of
having a lower secondary vs. a vocational degree)?
4.1 Vocational degrees: why are the estimates so close in the imputed compared to
the original histories?
Estimation results for the category ‘vocational degree’ in principle could be particularly
strongly affected by using imputed instead of original histories. This is because many
people gained another degree after having obtained a vocational degree first. This
applies to 27% of the respondents7 (Table 1). Altogether, 15% have more than one basic
vocational degree, 8% have a master craftswoman’s or technician’s degree in addition
to their basic degree, and 4% have both a university and a vocational degree. Thus, it is
of particular interest to see how exposure time and events are allocated into and away
from the category ‘vocational degree’ in the imputed histories as a consequence of the
respondents obtaining subsequent degrees at the same or a higher level, and because of
gaps between spells of educational enrollment.
The tables provided in the appendix map exactly to which education categories
exposure time and events from the original histories are allocated in the imputed
histories. Tables A1a - A1c do this for the imputation that was performed assuming that
respondents were asked for the first date they received their highest degree8. In Table
A1a in the appendix we can see that 99% of the exposure time from the original
5 Table A4 in the appendix gives standard errors for the estimates shown in Table 2. Here we can see that the
standard error for the estimate affected most by imputation, the estimate for the relative risk of first birth for
those with a lower secondary school degree as compared to a vocational degree, increases when using the
imputed histories. However, the estimates of this relative risk not only remain significant, they are significant
on a higher level in the imputed as compared to the original histories. This is because the size of the estimate
itself increases strongly enough to offset the increase in the standard error.
6 The reason these categories were included was mainly to avoid grouping them together with other categories
for which we do have a sufficient number of events. This way, effects for vocational, university, or lower
secondary degrees, for example, could be kept as clear-cut and generally comparable as possible.
7 given that the 4% of respondents with both a vocational and a university degree obtained their vocational
degree first.
8 The tables provided in the appendix only give cross-tabulations of exposure time and events for the first,
second, and fifth imputation. Cross-tabulations for the third and fourth imputation were omitted in order to
save space. General patterns of allocation of exposure time and events between the original histories and
imputed histories can be demonstrated sufficiently well on the basis of the three examples shown here.
http://www.demographic-research.org
149