Rotating Panel Design

Examples of Rotating Panel Designs

Under one type of rotation scheme, during each time period, one rotation group is canvassed for the first time, while another is canvassed for the final time. As an example, in the Labour Force Survey conducted by Statistics Canada, the sample of households is divided into six rotation groups, with a new group entering each month. Thus, in any month, one group is canvassed for the first time, one for the second time, and so forth. After 6 months of responding, households are retired from the sample. The Labour Force Survey conducted by the Australian Bureau of Statistics uses a similar scheme, dividing the sample into eight rotation groups; households are in sample for 8 months.

The rotation pattern can be more complex. The Current Population Survey, jointly sponsored by the U.S. Bureau of Labor Statistics and the U.S. Census [Page 772]Bureau to measure the U.S. labor force, uses eight rotation groups. A household is interviewed for 4 consecutive months, temporarily dropped from the sample for 8 months, brought back into sample for 4 more months, then retired. In any month one rotation group is interviewed for the first time, one for the second time, and so on, while one group is interviewed for the eighth and final time. For each of the surveys mentioned, the rotation pattern is said to be a “one-level” design because the respondents report for only the current period of time (1 month).

Other examples of surveys that use a rotating panel design are the National Crime Victimization Survey and the Consumer Expenditure Survey, each conducted by the U.S. federal government. In the Census Bureau's Survey of Income and Program Participation, although cross-sectional estimates are produced, the people in sample are contacted in a rotating pattern to obtain longitudinal estimates, that is, estimates of gross changes over time. These three surveys are examples of multi-level designs, as respondents report for several prior periods (months) during the same interview.

Under a different type of rotation scheme, groups cycle in and out of sample indefinitely or at least until the entire sample is retired. A good example is the former (before 1998) design of the monthly surveys of retail and wholesale trade in the United States, where the noncertainty sample was divided into three rotation groups. In each of 3 consecutive months, only one of the three groups was canvassed. This pattern was repeated, with each group reporting every third month, for the 5-year life of the sample.

...