Zabel: Imputed education histories and fertility analysis in the western German context
The types of bias named above will only be relevant if motherhood actually does
have a negative impact on progression to higher level degrees. In contexts where
motherhood and participation in education are very well compatible and motherhood
has no impact on subsequent educational attainment, then the types of bias described
above will not occur.
We have also found some evidence that first birth risks will be underestimated for
an education category if this category represents the highest level of education for many
respondents, but respondents nonetheless tend to frequently continue their studies or
reenroll in education after receiving this degree. As mentioned above, it is relatively
common in Germany to obtain a second vocational degree at the same level as the first.
If a questionnaire only asks for the first date the respondent ever obtained a degree at
their highest level, this appears to lead to a slight underestimation of first birth risks for
the category ‘vocational degree’ when imputations are based on this first date. This is
because respondents who subsequently enroll in their second vocational training spell
are mis-imputed as holding a vocational degree and not enrolled. Since risks of first
birth are typically very low while enrolled in education, this leads to a downward bias
for the ‘vocational degree’ category.
For further comparison, we also imputed education histories without using any
information on the date the respondents obtained their highest degree. For this
imputation, we used standard ages for the completion of respondents’ highest degrees.
Estimates based on this imputation gave more strongly biased results than the estimates
based on imputations that did make use of the date the highest degree was obtained.
Thus, it seems that when it is not possible to collect complete education histories in
surveys, collecting at least the date respondents obtained their highest degree and not
only the level can improve the quality of imputations.
Altogether, when we did make use of the date the highest degree was obtained,
imputation of education histories did not cause much bias, with the exception of first
birth risks for the category ‘lower secondary degree’ ('Hauptschulabschluβ'). Estimates
based on imputed histories may become more strongly biased for future cohorts than for
the cohort of 1964 studied here if it becomes more common to obtain higher level
degrees after having already completed a lower level degree some time ago. This might
occur for Germany if more women obtain a master craftswoman’s degree (a higher
level vocational degree) sometime later in their life course. Also, if people begin to take
breaks between completing a bachelor’s degree and beginning their studies for a
master’s degree, this might also make imputations more problematic. International
bachelor’s and master’s degrees have only very recently been introduced in Germany.
However, imputations of education histories in countries where they are already well
established may be more problematic. In countries where it is generally more common
162
http://www.demographic-research.org