Implicit stratification through spatially balanced sampling designs and their applications in official statistics

Abstract

Spatially balanced sampling methods have traditionally been applied in
environmental and forestry surveys, and some of these designs have also
found uses in official statistics. When the Horvitz–Thompson estimator is
employed to estimate population parameters, such methods demonstrate high
efficiency in the presence of positive spatial autocorrelation of the variable
of interest, that is a common feature of many socio-economic variables like
income. Spatial autocorrelation can produce effects similar to clustering,
where neighbouring units provide similar information. To mitigate this is-
sue, it is essential to ensure that the selected sample units are spatially well-
distributed. This paper pursues two main objectives. First, it reviews some
spatially balanced sampling designs used in official statistics and describes
some examples of their application. Second, it shows that these designs can
serve as an alternative to classical stratification with proportional allocation
to the number of units. Finally, the paper discusses the advantages and limi-
tations of using spatially balanced sampling in official statistics.

More information

Identifiers

ORCID for Francesco Pantalone: orcid.org/0000-0002-7943-7007

ORCID for Paul A. Smith: orcid.org/0000-0001-5337-2746

Catalogue record

Export record