The LFS Design

The LFS design must meet many Labour Force Survey.

• It must provide for efficient estimates of month to month changes in levels of employment and unemployment.
  (This implies the use of common samples, from one month to the next.)
• It must ensure the respondent load is kept at a reasonable level.
  (This implies rotation of households out of the sample, after a reasonable period.)
• It must be operationally straight-forward to operate, both minimising the risks of inadvertent error due to complexity, and providing timely delivery of results.
• Finally, it must be cost-effective; there should be no other viable design that could produce estimates having lower error, for the same or lower cost.

To maximise the efficiency of the survey, both costs and variability need to be taken into account. Factors contributing to the overall costs are ...

• the sample size;
• length of interviews;
• travel time between interviewer's home and the sampled households;
• the travel time from one sampled house to another;
• the number of interviewers;
• training and infrastructure costs per interviewer.

Cost-efficient designs are achieved by clustering together the houses being sampled; thereby minimising travel costs. However as clusters of houses tend to be more alike, in terms of social and economic characteristics, than houses from different locations, then an unfortunate consequence of this is that clustering increases sampling-error.

Each five years, as updated information is obtained from the population-census, the survey is redesigned. The cost structure for travel and interviewing is re-calculated, taking account of the intended approach to enumeration. Also the variance structure is re-modelled, to provide an optimal design. Figure 1, on the next page, shows the relationship between `per block costs' and `per household costs', and the optimal cluster-size for a given region, taking account of the variance components of the series.