In the CPS, households are interviewed for four consecutive months, drop out of the sample for the next 8 months, and then are interviewed again for the same 4 calendar months a year later, before retiring permanently from the sample. This practice avoids placing too heavy a burden on the households selected for the sample, but improves measures of change. Seventy-five percent of the sample remains the same from month to month, and 50 percent from year to year. The overlap of this 4-8-4 rotation scheme helps to improve CPS estimates over time, since the sample's continuity results in a positive correlation between estimates from different months.

Despite this continuity, individual rotation groups are subject to different survey conditions. An empirical illustration of the effects of time in sample can be seen in the graph below. The greater proportion of nonresponse in month-in-samples (MIS) 1 and 5 are likely due to actual differences in response probabilities for these groups. Such differences could arise for a variety of reasons, including mode of interview. With MIS 2-4 and 6-8 households, data collection is primarily by telephone through computer-assisted interviewing; for MIS 1 and 5 household data are typically collected by Census field representatives making personal visits. Interviewers attempting to contact households for the first time (or, when returning to the sample in MIS 5, for the first time in eight months) might be less successful at finding someone at home, or have a harder time (re)establishing rapport once the household is contacted. Thus, MIS 2-4 and 6-8 typically have lower nonresponse rates than months 1 and 5. Nonresponse that varies by month would affect estimates if uncorrected, but actual CPS estimates are weighted to account for nonresponse (for further discussion of nonresponse, see Nonresponse Rates or Tech Paper 77, Chapter 2-3).