Zabel: Imputed education histories and fertility analysis in the western German context
attaining any degree. After this time, up until the time of interview, she is considered to
have no degree and not to be enrolled.
The above description applied to the first four imputations, where we assume that
we have information on the date the respondent received her highest degree. For the
fifth and last imputation, however, we assume that the survey contains no information
on the date the respondent received her degree, and only provides information on the
level of that degree. Here then, in addition to imputing the date the respondent switched
from school to vocational or university education, we also need to impute the date the
respondent received her vocational or university degree. We imputed the date the
respondent received a vocational degree as July, three years after completing school.
Vocational training can take between two and four years, so three years was chosen as a
intermediate duration. The date a respondent received a university degree was imputed
as July, six years after finishing school. Six years is a realistic duration for many
subjects, although study durations can vary greatly in either direction.
If the respondent’s highest degree was a school degree, in the fifth imputation
where we do not know the date the respondent received this degree, we need to
differentiate between lower secondary school degrees, advanced lower secondary
degrees, and upper secondary degrees. For lower secondary degrees, we assume the
respondent attended school up to ninth grade, which corresponds to July the year she
turned 15 or 16, depending on her birth month. For advanced lower secondary degrees,
we impute July the year the respondent turned 16 or 17, and for upper secondary
degrees, July the year the respondent turned 19 or 20.
The next section compares estimation results for the effect of education on risks of
first birth using the complete and imputed histories, and discusses explanations for
deviations.
4. Results
The results in Table 2 show that the first four imputations generally produce estimates
that are quite close to the ones we obtain using the original histories. In terms of
estimates for the baseline, deviations from the original histories are quite small in each
of the four cases. Estimates of the effect of having a vocational as compared to a
university degree are likewise very close to the estimate produced using the original
histories. The strongest deviation between the results obtained from the imputed as
compared to the original histories, however, occurs for the relative risk of first birth for
having a lower secondary vs. a vocational degree. The original histories produce an
estimate of 1.52 for the relative risk of first birth for having a lower secondary degree as
compared to the reference category. The first four imputed histories, on the other hand,
148
http://www.demographic-research.org