The Labour Force Survey sample can be thought of as comprising eight sub-samples (or rotation groups), with each sub-sample remaining in the survey for eight months, and one rotation group "rotating out" each month and being replaced by a new group "rotating in". This sample rotation is important in ensuring that seven-eighths of the sample are common from one month to the next, to ensure that changes in the estimates reflect real changes in the labour market, rather than the sample. In addition, the replacement sample is generally selected from the same geographic areas as the outgoing one, as part of a representative sampling approach.
When considering movements in the original estimates, it is possible to decompose the sample into three components:
- the matched common sample (survey respondents who responded in both July and August;
- the unmatched common sample (survey respondents who responded in August but who did not respond in July, or vice versa); and
- the incoming rotation group (survey respondents who replaced respondents who rotated out in July).
The detailed decomposition of each of these movements is included in the data cube 'Insights From the Original Data'.
In considering the three components of the sample, it is important to remember that the matched common sample describes the change observed for the same respondents in July and August, while the other two components reflect differences between the aggregate labour force status of different groups of people.
While the rotation groups are designed to be representative of the population, the outgoing and incoming rotation groups will almost always have somewhat different characteristics, as a result of the groups representing a sample of different households and people. The design of the survey, including the weighting and estimation processes, ensures that these differences are generally relatively minor and seeks to ensure that differences in characteristics of rotation groups do not affect the representativeness of the survey and its estimates. Monthly estimates are always designed to be representative of their respective months, regardless of the relative contribution of the three components of the sample.