Demographic Research: Volume 21, Article 6
overestimation, would no longer be counteracted by this second source of imprecision
in the imputed histories.
Altogether, although the bias for the ‘lower secondary school degree’ category is
the greatest of all categories examined here, the total number of events involved is
moderate. Teenage pregnancy rates are not very high in Germany, not many people
have children at a very young age before beginning or while still in vocational training.
In other contexts, this may play a stronger part. Whether estimates of first birth rates for
school degree categories will be biased when using imputed histories depends on
whether people actually do drop out of vocational training or decide never to start
vocational training if they become pregnant, leaving them with their school degree as
their highest educational degree. In some countries, parenthood and vocational training
may be compatible. This would lead to a lower bias when using imputed histories, even
if pregnancy rates among those in vocational training or among those who have not yet
begun vocational training are high. A similar situation may arise in contexts where it is
common to drop out of school after becoming pregnant. In Germany, this does not seem
to be a major issue, but in other countries it may lead to an overestimation of the risk of
first birth for the ‘no degree’ category when using imputed histories.
4.4 Imputation without knowledge of the date the highest degree was obtained
The fifth and last model estimate shown in Table 2 uses education histories that were
imputed without making use of any information on the date a respondent’s highest
degree was obtained. Here, we simulated the case where the questionnaire asks for the
respondent’s highest degree only, and does not additionally ask for the date this degree
was obtained. Using this imputation, the estimates for the baseline diverge more
strongly from the estimates using the original histories than is the case for the first four
model estimates based on imputed histories that do make use of information on the date
the highest degree was obtained (Table 2). For the last imputation, it was assumed that
respondents obtained their degrees after standard durations of enrollment, as described
in the methods section. Since many respondents are likely to have taken longer than a
given standard duration of time to obtain their degree, or did not enroll in vocational or
university education straight after school, or obtained more than one degree, the
procedure used for the last imputation is very likely to extensively miscode exposure
time that was actually spent in vocational or university education as exposure time
spent holding a vocational or university degree. As we can see in Table A3a, this is
indeed the case. Of the time originally spent enrolled in vocational training, 34% is
imputed as time spent holding a vocational degree and not enrolled in education. Seen
from a different perspective, of the imputed exposure time for the category ‘vocational
http://www.demographic-research.org
159