Do imputed education histories provide satisfactory results in fertility analysis in the Western German context?

Demographic Research: Volume 21, Article 6

In addition, further births are allocated into this category from the categories ‘in
vocational training’ and ‘in school.’ The number of births added to this category is
disproportionately high compared to the amount of additional exposure time (Table
A1b and Table A1c). Thus, it seems that some people dropped out of vocational
training or further school education after becoming pregnant without completing their
vocational degree or higher level school degree¹¹. The lower secondary school degree
then remains their highest degree, and the time during which they were enrolled in
vocational training or higher level school education is not registered in the imputed
histories. They are therefore represented as having become pregnant while they were
not enrolled and were holding a lower secondary school degree, although in reality,
they became pregnant while enrolled in vocational training or higher level school
education. These two factors - people who did not have children being selectively
removed from the ‘lower secondary degree’ category, and people who did have children
being added to this category - seem to have caused the overestimation of first birth
risks for the category ‘lower secondary degree’ in the imputed histories¹².

The size of the bias may even be underestimated here. This is because the original
histories may give slightly biased results themselves. In the form they were distributed,
the original histories were pre-prepared so as to close small gaps of up to four months
between school spells or between school and subsequent activity spells (like vocational
training spells) (Hillmert et al. 2004, part IV, p.24). Just like in the imputed histories,
this has the potential to lead to an overestimation of first birth risks for school degree
categories. Thus, differences between the original and imputed histories only arise if the
gap between school and vocational training is more than four months long. If small
gaps of up to four months had not been filled in the original data, then the difference in
estimates between the imputed and original histories would likely have been even
greater¹³.

Although the estimate for the category ‘lower secondary degree’ was quite
strongly biased in the imputed histories, not much bias was found for the other types of

¹¹ As described in the methods section, the date of first birth was backdated by nine months. Thus, strictly
speaking, the dependent variable is always the rate of transition to first pregnancy. However, we have often
used the term ‘first birth’ since this is more customary.

¹² In the case of the ‘lower secondary degree’ category, the imputed amount of exposure time and events does
not differ between the first and second type of imputation shown in the appendix. This is because it is
generally not possible to receive the same school degree twice, thus there is no difference between the first
and last date the highest degree was obtained.

¹³ The data preparation procedure used for the originally distributed data closes small gaps of up to four
months between school and vocational training spells by extending the length of the school spells. In the
imputations used for the present analyses, gaps between school and vocational training spells are effectively
closed by extending the length of the vocational training spell backwards up to the point in time school
education is assumed to have ended. Both procedures, however, in the same way lead to an overestimation of
first birth risks for school degree categories.

http://www.demographic-research.org

<<   22   23   24   25   26   27   28   >>

More intriguing information