Demographic Research: Volume 21, Article 6
5. Conclusion
This study has investigated whether imputed education histories can act as satisfactory
substitutes for complete education histories when analyzing first birth transitions, in the
case of western Germany. Our results do not indicate very extensive distortions of
model estimates when using imputed histories. This appears to be related to the special
nature of educational trajectories in Germany. Nonetheless, small to moderate
distortions did occur, and examining how exposure time and events from the original
histories were allocated in the imputed histories has enabled us to identify some general
patterns of misallocations.
One finding is that imputations seem to become problematic whenever there are
breaks between education spells during which people hold lower level degrees for some
amount of time before reenrolling or continuing their studies to obtain a higher level
degree. The greatest distortion of this nature that we found was for the category ‘lower
secondary degree’ ('Hauptschulabschluβ'). The imputed histories gave upwardly biased
results for this category. This is because childless women were more likely to continue
their studies and attain a post-secondary degree. Any gap during which they held a
lower secondary degree before starting their post-secondary education was
misrepresented as a consequence of imputation.
A slight upward bias was also found for the category ‘vocational degree’ for one
specific type of imputation. For that imputation, we simulated a case where the
questionnaire asks only for the last date the respondent ever obtained a degree at their
highest level. A distinction between the first and last date respondents received their
highest degree can be relevant in the German context, because it is common to have
more than one vocational degree at the same level. However, the bias did not turn out to
be very large. It seems that not enough people obtained a second degree late enough in
their life course to lead to a serious extent of misallocation of exposure time in the
imputed histories.
Thus, gaps between education spells are one source of difficulties for imputation.
A further general pattern of misallocation of exposure time in the imputed histories is
related to cases where people drop out of education upon having a child. In these cases,
births that in reality took place while enrolled in education will be shifted to a lower
degree category in the imputed histories. This will also bias estimates of first birth risks
for lower educational degrees upwards. For the cohort studied here, this was shown to
occur for people enrolled in vocational training who had previously completed a lower
secondary degree ('Hauptschulabschluβ'). This factor further contributed to the
overestimation of first birth risks for the ‘lower secondary degree’ category when using
the imputed histories.
http://www.demographic-research.org
161