Addresses are selected on a random basis from the Postcode Address File (PAF) to take part in the Active Lives survey. At each address, up to two adults can take part in the survey.
Two unique login codes for the online survey are provided on each invitation, and at the paper questionnaire mailing (M3), up to two questionnaires are sent (if one adult in a household has already completed the questionnaire online, then only the unused password and one questionnaire will be sent in reminders, which follow the initial invitation).
Each adult who completes the questionnaire can claim a £5 paper voucher/or online gift code for doing so. This methodological approach was used in preliminary testing by Sport England, prior to commissioning Ipsos to undertake the Active Lives survey.
In the preliminary testing work on this approach carried out by Sport England, four log-in codes were provided in the invitation letter (each with a conditional incentive).
While this removed the need for any selection of adults in nearly all households (only 1% of households have more than four adult residents), it has the potential for fraud whereby a household can complete the survey for more adults than actually live there in order to obtain the incentives on offer.
The work for Sport England found that 4% of addresses filled in the survey for more adults than they listed as living at the address, and the average size of participating households was higher than would be expected (2.19 compared to an average household size of around 1.41).
Collecting up to two completed questionnaires per household reduces the chance of receiving fraudulent responses. There is still scope for fraud using this approach, but this is reduced. Therefore, for the main Active Lives adult survey, two logins were provided on each letter.
Process for selecting adults within a household
There are no selection criteria to taking part in the survey. Instead, any two members of the household (aged 16+) can take part.
Using this approach, it is estimated that 93% of the sample are the ones that would be selected using a quasi-random approach (compared to 57% of the sample if only one adult was selected).
There are many approaches that could be taken to selecting adults within households. For instance, the two adults with the most recent birthdays, or the adults with the two next birthdays, could be selected to complete the questionnaire.
These are commonly referred to as quasi-random approaches as they are roughly equivalent to a fully random approach.
While this would randomise the selection process to a degree in households where there are more than two adults, in self-administered surveys it just adds another barrier to completing the survey and has shown to not be carried out correctly in about 20% to 25% of cases.
Impact of up to two adults per household taking part in the survey
There is a small impact on data from obtaining more than one response per household, which results from the increased clustering effects – we would expect people in the same household to have on average similar levels of sport and physical activity, which would result in a small loss of precision.
However, this loss in precision is more than outweighed by the gain in precision from the resulting lower variability within household selection weights.
For any estimates that are reported by sex, then the clustering effect within the household is typically negligible as for the vast majority of households with more than one respondent, one will be male and one female.