About this Release
The MPS adopts a multi-stage design in which the sample is clustered within first stage units. A high level of clustering will reduce costs in travel between first stage units which account for a high proportion of overall costs. However this will result in higher variances as fewer first stage units will be selected (for a fixed total sample size). On the other hand a low level of clustering will cost more in travel but will produce lower variances on estimates.
The objective of the optimisation process is to determine the level of clustering that achieves the best trade-off between costs and variance by minimising total cost for a fixed level of accuracy. Key components of the optimisation process are the cost and variance models which provide the link between sample sizes at each stage of selection and resulting costs and variances, respectively. As survey accuracy deteriorates over the life of a design, the prime objective of the sample redesign is to return to the level of accuracy achieved at the beginning of the current (1996) design period.