摘要：The estimation of Austrian unemployment rates is based on data of
the labour force survey (LFS). It is possible to calculate direct, design based
estimates with fixed precision for population subgroups for which the sample
size is known due to the sampling design.
Sometimes we are interested to estimate unemployment rates for population
subgroups in which the sample size is random and often small. In this paper
we conducted a simulation study to compare the performance of several
model-assisted estimation methods with the direct estimator using auxiliary
information from an administrative source when estimating unemployment
rates for population subgroups. The results showed that if sample-sizes in
a subgroup are small, model-based estimators outperform the design-based
estimator.
