Zabel: Imputed education histories and fertility analysis in the western German context
school degrees, that is, for the categories ‘advanced lower secondary degree’ and ‘upper
secondary degree’ (Table 2). As we can see in the appendix, this is the case despite a
large extent of misallocation of exposure time for these categories. Only 49% and 33%,
respectively, of the original exposure time is correctly allocated to these school degree
categories. People who gain an upper secondary school degree, on the basis of which
they can apply for university, especially tend to take long breaks before enrolling at a
university (or in vocational training). They therefore in reality spend a lot of time
holding an upper secondary school degree without being enrolled, which in the imputed
histories is represented as a time of enrollment in university education (or vocational
training), for all of those who do in the end receive a university (or vocational) degree.
Distortions of the model estimates for the school degree categories ‘advanced lower
secondary degree’ and ‘upper secondary degree’ do not seem to be dramatic. However,
the number of events for these categories is too small to draw any general conclusions.
Larger sample sizes could show whether the uptake of post-secondary education is
influenced by motherhood for these school degree categories or not, and whether or not
imputation would therefore lead to distortions.
A detail of the imputation procedure used here may have also contributed to
misallocation of exposure time. As described in the methods section, for people without
a post-secondary degree, a short period of post-secondary education was imputed
nonetheless. The idea was that people who had not completed a post-secondary degree
may still have begun post-secondary education, but have dropped out without
completing a degree. Only a short period of one or two years, depending on
respondents’ type of school degree, was imputed to account for this. This may have still
overstated average lengths of enrollment in post-secondary education for respondents
without a post-secondary degree though, since many may have never begun any post-
secondary education at all. A reason that some respondents never entered post-
secondary education may be that they had a child. Exposure time which these
respondents in reality spent holding a school degree is allocated away from the school
degree categories as a consequence of this imputation procedure. This is likely to bias
the first birth estimates for school degree categories downward when using the imputed
histories. This source of mis-imputation will thus work in the opposite direction to the
source of mis-imputation discussed above, which worked towards overestimating first
birth risks for school degree categories. If lengths of enrollment in post-secondary
education for those without any post-secondary degree are indeed overstated by the
imputation procedure used here, and if a method could be found to more realistically
represent these interrupted phases of post-secondary education in the imputed histories,
then, in sum, the overestimation of first birth risks for school degree categories would
be even greater. This is because the first source of mis-imputation, which led to an
158
http://www.demographic-research.org