Page 17 - Invited Paper Session (IPS) - Volume 1

IPS30 Stefan D. et al.
3.2 Lessons learned of the 2011 census
Retrospectively, the modular concept of the German census 2011
combining register use and primary statistical data collections was successful.
A qualitative evaluation came to the result that the model worked well, and
(while identifying proposals for an improved implementation) should be the
basis for the census 2021. At the same time, it was concluded that the quality
of the population registers was still not considered sufficient for the purposes
of a census so that the supplementary household sample survey was
maintained as an element to correct for over- and under-coverage in the
registers (and to collect the data for variables not covered by the registers).
Against this background, the household sample survey remains an integral
elemental part with an opitimization of the model considering census 2011
results. The model was scientifically accepted and achieved a high precision.
At the same time the model reduced the burden and the cost on data
collection for the statistical offices and the respondents compared to the
former traditional census considerably. The household generating procedure
was another new element which turned out for being a precise way to
determine family and household data down to the local level.

3.3 current objectives for the 2021 census
Changes for the census 2021 consider census 2011 experiences in quality
aspects and possibilities to reduce complexity and thus to be more timely with
results. Additionally, modifications aim to make the results easier to
understand and raise general acceptance of the model. The main amendments
for the census 2021 are:
• The supplementary household sample survey used to adjust the register data
in municipalities with 10,000 inhabitants or more will be carried out in all
municipalities in 2021. An analysis of the 2011 data showed that the special
survey on addresses with implausibilites in smaller municipalities did not fulfil
the adjustment needed. It turned out that the overall assumption of a better
register quality in smaller municipalities was correct, but the correcting rate
was still higher than assumed before. To avoid a large scale of small
municipalites with high sample rates by stretching the survey concept and the
relative error margin of big municipalities 1:1, the relative error margin will be
switched to an absolute error margin using a target function for the
municipality size in between. Secondly, for small cities using the same
administration for their population registers the sample survey will be drawn
on the joined municipality level. The results on this level will be splitted for
each municipality afterwards using a two-step estimation model with survey
results and register data.
• The interaction of the different census components need to be designed early
on to allow a comprehensive technical approach that liaise the individual parts

6 | I S I W S C 2 0 1 9