Investigating the Cube Sampling Method for Household Surveys
The use of information in the efficient design of surveys has been studied extensively. Well known methods include stratification and probability proportional-to-size sampling. These methods are designed to select efficient samples when there is only one survey characteristic of interest. More recently Deville and Tille (2004) developed the cube sampling method with the potential to select efficient samples when there are multiple characteristics of interest. Specifically, the cube method selects a balanced sample on a set of design variables. A balanced design has the property that the Horvitz–Thompson estimates of total for the set of design variables equal their known totals. If the design variables are well correlated with the survey characteristics of interest then a balanced sample will be efficient (i.e. less cost for the same level of standard error on the survey estimates).
The Household Survey Methodology section recently undertook a preliminary study to measure the reduction in standard errors of using the cube method for the selection of ABS household surveys. ABS household surveys are area samples, typically with Census Collection District (CD) as the Primary Sampling Unit (PSU). The study measured the impact on the standard error of balancing the sample of CDs on a set of CD-level design variables, obtained from the Census. The preliminary results of the study suggest that cube sampling has the potential to provide significant cost savings, particularly for the Labour Force Survey.
With the availability of meshblocks, which is a much smaller geographic unit than the CD, future ABS household surveys may use this as the PSU. The study found that selecting a balanced sample of meshblocks, rather than of CDs, would provide further cost savings.
In future, the study will consider the complexity of maintaining a balanced sample over time while allowing for sample rotation, as well as operational and implementation issues.
For more information contact James Chipperfield on (02) 6252 7301.