Within the framework of the quality control and evaluation program for
the Portuguese 2001 Census, the Portuguese statistical office (INE)
conducted a Post Enumeration Survey (PES) to measure quality. The main
aims of the PES were to evaluate coverage errors and content errors for
the target populations. The PES is a probabilistic sampling survey
representative at NUTS II level. This paper describes the methodology
for this survey. The paper includes a discussion of sample size and
allocation resulting from the imposition of maximum coefficients of
variation for a set of variables both at regional and national level.
The methodology used to obtain predictions for resident populations and
dwellings is also presented. These predictions
are used in the definition of inclusion probabilities for the primary
sampling units. The sampling design is finally compared with two
alternative designs (with a smaller number of stages), concluding for
the advantage of the proposed design in regard to the survey goals.