Zabel: Imputed education histories and fertility analysis in the western German context
one is employed (Kerckhoff 2001). The Swedish educational system is similarly
flexible, with a wide range of policies aimed at supporting students who reenter
education at a later age. A very high proportion of students in Sweden resume education
after already having gathered labor market experience, or take part-time courses while
employed (Hoem, Neyer, and Andersson 2006). We suspect that it would be difficult to
approximate education histories using only sparse information for countries such as
Sweden, Great Britain, or the United States.
The reason why imputation may be difficult in these countries is that there can be a
great deal of variation in the timing of educational reentry. It may be very difficult to
infer at what point in time people resume education exactly, or whether they transferred
directly from school to post-secondary education. In countries with more flexible
education systems, more students may also reenroll to acquire a further degree in a
different subject but at the same level as before. This can also cause difficulties for
imputation, since surveys generally only collect one date for respondents’ highest
degree. If, following our imputation method, we assume that people are enrolled in
education up until the time they obtain their highest degree, substantial parts of
respondents’ activity histories are likely to be misrepresented. For instance, people will
be coded as being enrolled in education for long periods of time during which they are
actually employed or pursuing some other activity before obtaining their highest degree.
Another concern is that they will be imputed as not enrolled after they obtained their
highest degree, even if they do subsequently continue or reenter education.
Confusing time spent enrolled and not enrolled in education is especially
problematic in the context of fertility analysis. In many studies, educational enrollment
has been shown to have an extremely strong negative impact on rates of transition to
first birth (e.g. in Rindfuss, Morgan, and Swicegood (1988), Blossfeld and Huinink
(1991), Kreyenfeld (2008), Vikat (2004), and Edwards (2002)). Using imputed
education histories may therefore become problematic if a lot of people continue their
education without completing a further degree, or obtain a lower level degree after
having already attained their highest degree, or complete a further degree at the same
level as their previously completed highest degree. Since imputations are based on
respondents’ highest degree alone, any time of enrollment after they completed their
highest degree is not registered. Respondents are instead imputed as not enrolled and
holding their highest degree. However, as mentioned above, risks of first birth tend to
be substantially lower for women who are in reality currently enrolled. Therefore, risks
of first birth would be underestimated for non-enrollment categories when using the
imputed histories. To some extent, this will be a problem in the context of any country,
as some people will always continue their studies for some time without completing a
further degree. However, it is likely to be especially relevant in countries that are
supportive of students wishing to obtain a degree at the same level but in a different
138
http://www.demographic-research.org