The effects of selecting multiple respondents per household for a survey of people in paid work: A statistical and cost assessment

For PAF based surveys such as the Skills and Employment Survey a decision is needed on the number of working adults to select per household. In the past the SES has always selected just one, with the selection being at random from all working adults in the household. But other options are to select all working adults, or to apply a cap and select up to, say, two. The decision is not straightforward, but we have conducted analysis, described in this note, that suggests that selecting more than one adult per household would be a sensible change for the SES.

In addition to the statistical benefits of such a change, selecting multiple adults per household should provide fieldwork efficiencies that will enable an increase in the number of interviews that can be achieved within a fixed budget. We provide an estimate of the scale of the increase possible.

A further consideration for a long-running study with a valuable timeseries is that of comparability over time. There remains a concern that the nature of non-response bias is affected by a change in the selection approach. This is difficult to assess ahead of fieldwork. However, we argue this uncertainty is set against the more significant risks to changing response bias presented by a context of falling response across surveys in general which the post-pandemic situation may have exacerbated.