Online Program

Efficient sample design for an establishment survey

Keywords: Sample design, design effect, stratification, Synar

All States have been required for over 15 years to perform annual inspections of tobacco retail outlets to assess the rate of illegal sales of tobacco products to minors. Furthermore, the federal regulation (Synar amendment) requires that these inspections be carried out on a random sample of outlets. With all States providing data for over 10 years, we have now accumulated a wealth of data that allows for detailed study of factors underlying efficient sample design. This presentation compares the design effects for different States, identifying the sampling designs that produced design effects close to 1 (or less than 1). There is some evidence to indicate that clustered design in some cases can actually be more efficient than simple random sample designs. Stratification is an integral part of most States’ surveys and we measure the extent to which this sample design feature helps to reduce precision. Finally, given the recent and increasing pressure to reduce costs, we identify those design features that have reduced costs without unduly affecting the precision of the results.