All Times ET
Program is Subject to Change
Sample Design in Establishment Surveys (308404)*Paul Anthony Smith, University of Southampton
Establishment surveys are characterised by skewed populations, with a few large units, with a disproportionate contribution to statistics, and a large number of smaller units. They also have good auxiliary data which can be used for efficient sampling and estimation. We discuss the benefits and challenges of employing such information, typically derived from a business register. Since the staple approach to sampling for such populations is stratification, we summarise the properties of stratified sampling, including procedures for determining the number of strata and their definition, and allocating the sample according to variance or sample size constraints. We consider challenging topics for which good auxiliary data are not available, and how sampling for such characteristics may be supported by new forms of data.
We also summarise other sampling approaches in business surveys, including cut-off sampling, PPS and indirect sampling, and give examples of their operation.