The main section of the questionnaire is designed to gather information on which activities people have taken part in, how much time has been spent doing them and the level of intensity.
This information can then be used to create measures which correspond to the KPIs in the Get Active strategy. For this reason, the key reference periods used are the last year and the last 28 days since these are the reference periods in the KPIs.
It should be noted that there are recall errors associated with long reference periods for these types of behaviour. Errors associated with use of a 28-day and a year-long reference period include telescoping (where people bring in activities which actually took place earlier) and omission of activities which took place in the period because they have been forgotten.
In gathering detailed activity data for a 28-day period, some respondents will have multiplied what they did in the last week by four, which may have led to errors in the calculation, or in reporting information for the 28-day period which was not strictly correct if not all four weeks were like the most recent week.
In gathering information on participation in activities, respondents were not required to select ‘no’ for activities they had not done. Given the volume of activities to select from, this would have been an excessive burden on respondents.
This means that in analysing activity and deriving the key measures, any activities which have not been ticked are regarded as not having been done. This contrasts with other questions, where the absence of a tick is treated as missing data.
An important measure for analysis of the activity data is socio-economic status. The key measure used is NS-SEC (National Statistics Socio-Economic Classification). This is derived from information about people’s occupations and their workplace.
In other official surveys, this is derived using detailed questions about job description, qualifications needed to do the job, the place where they work as well as information on number of employees and supervisory responsibilities which is then coded to SOC (Standard Occupational Classification) which is in turn used to derive NS-SEC.
However, this full approach to the collection of the data used to code SOC and then NS-SEC relies on interviewer administration (in person or by telephone).
The Office for National Statistics (ONS) have created a short self-coded version which can be used to create five classes of NS-SEC. This is not as accurate as the interviewer-coded counterpart, with 75% agreement with the full version.
ONS ran an experiment in 2001 on their Omnibus, comparing interview-coded NS-SEC (the measure collected in the Labour Force Survey) against self-coded NS-SEC (the measure collected in the Active Lives Survey) for the same participants.
They found that there was a tendency for people to overstate their NS-SEC level, with a lower proportion coding themselves as semi-routine and routine than in the full version.
Nonetheless, because of the importance of socio-economic differences in participation and the under-representation of some groups in the survey responses, NS-SEC has been used for weighting, even though it is not directly comparable with the NS-SEC version in the Annual Population Survey, which provides the targets to which we weight the data.
To mitigate the impact of this, highest educational qualification, which is more comparable, between the Active Lives survey and the Annual Population Survey has also been included in the weighting. The weighting is explained in more detail in the Weighting section.
In the dataset, where follow-up questions have only been asked for a subset of respondents, the data have been used to create derived variables which are re-based to the population. For example, volunteering more than once in the last year, and number of long-term limiting disabilities.