7 Appendices
Methodology
7.1 2,010 Computer Assisted Telephone Interviews were completed in Scotland between 8 th September and 14 th December 2005 by IFF Research Ltd. from its Telephone Interviewing Centre in London. In addition to these, a further 15,696 interviews were conducted in England 31.
7.2 A Random Digit Dial ( RDD) sampling method was chosen in order to gain a sample representative of the population of England and Scotland of broadly working age ( i.e. aged 16-64). This involves several stages:
- A sample of random numbers was generated by computer by UK Changes, a specialist telephone number generation and look-up agency
- The random numbers were generated according to known area codes (Subscriber Trunk Dialling - STD) in proportion to the number of interviews required in each Government Office Region ( GOR). The sample size per region varied according to whether individual RDAs or other organisations wished to fund additional interviews.
- Telephone numbers were then 'pinged'. This procedure involves an autodialler ringing each of the numbers for a split second (a very quick 'ping' can be heard from the receivers, hence the name) in order to establish that there is a connection. Therefore, unallocated telephone numbers are removed from the sample
- Normally at this stage in a household survey the ensuing numbers are then screened against a national database of business numbers in order to remove these. However, for the SBS Household Survey this was not done, because there is a chance that a respondent, especially one that is self-employed, might use the same number for both business and household calls
- The procedure produces a random sample of households, but not of individuals. Therefore for two-thirds of the calls, on making contact by telephone, the person in the household aged 16-64 with the next birthday coming up was requested
- Experience has told us that young people, and especially young males, are harder to reach via this method, and more likely to not want to take part in the survey. Therefore, for one third of all calls anybody aged 16-24 in the household was requested in the first instance. If nobody in the household was in this age group, or if the 16-24 year old was unavailable or did not want to do the interview, then the (next) person with the next birthday coming up was interviewed instead
- Despite this attempt to improve the representation for younger people, it was still necessary to 'upweight' 16-24 year olds at the analysis stage
7.3 IFF interviewers called respondents between 5 pm and 9 pm during weekdays, 10 am and 5 pm during weekends. Potential respondents unable to do the interview there and then would be called back at a convenient time. Each piece of sample was called up to a maximum of eight times before retirement.
7.4 Within England only, areas known to have high proportions of ethnic minorities were oversampled.
7.5 Because of this oversampling by region and ethnic group, and also because of sample bias against young people, at the analysis stage weighting was required to make the data representative of the population of England/Scotland aged 16-64.
7.6 Weighting targets were drawn from the 2001 census. Although more recent estimates exist in terms of population by region, age and gender, detailed recent information on ethnicity does not exist, hence the need to use older data. Weighting targets were imposed within each region, using an interlocked grid which combined age, gender and ethnic group.
7.7 Within Scotland the data were weighted according to an interlocked grid combining age, gender and ethnicity:
- interlocked age x gender (8 cells: 16-24 male, 16-24 female, 25-34 male, 25-34 female, 35-44 male, 35-44 female, 45-64 male, 45-64 female) x ethnicity (2 cells: {i} White British/White Irish/Mixed Race; {ii} other)