The Scottish Government

« Previous | Contents | Next »

Listen

CHAPTER 12 APPENDIX A: TECHNICAL NOTES

12.1 Sampling

12.1.1 Sample design

The survey is based on two cohorts of children: the first aged approximately 22.5 months at the time of the sweep 2 interview and the second aged approximately 46.5 months. Initially, a named sample of approximately 13,000 children was selected from the Child Benefit Records to give an achieved sample at sweep 1 of 8076 cases.

The area-level sampling frame was created by aggregating Data Zones. Data Zones are small geographical output areas created for the Scottish Government. They were used to release data from the Census 2001 and are used by Scottish Neighbourhood Statistics to release small area statistics. The Data Zone geography covers the whole of Scotland. The geography is hierarchical, with Data Zones nested within Local Authority boundaries. Each Data Zone contains between 500 and 1,000 household residents. More information can be found on the Scottish Neighbourhood Statistics website: http://www.sns.gov.uk.

The Data Zones were aggregated to give an average of 57 births per area per year (based on the average number of births in each Data Zone for the preceding 3 years). It was estimated that this number per area would provide enough issued cases to achieve a sample of 8,000 (this is reliant on the birth rate remaining roughly constant). Once the merging task was complete, the list of aggregated areas was sorted by Local Authority ³⁶ and then by the Scottish Index of Multiple Deprivation score. One hundred and thirty areas were then selected at random. The Department of Work and Pensions ( DWP) then sampled children from these 130 sample points.

Within each sample point, the Child Benefit Records were used to identify all babies and three-fifths of toddlers who met the date of birth criteria (see Table 12.1). The sampling of children was carried out on a month-by-month basis in order to ensure that the sample was as complete and accurate as possible at time of interview.

In cases where there was more than one eligible child in the selected household, one child was selected at random. If the children were twins they had an equal chance of being selected. If the eligible children within the same household were in different age cohorts the baby had a higher chance of being selected, this was to ensure the ratio of babies to children remained constant.

After selecting the eligible children, the DWP made a number of exclusions before transferring the sample details to ScotCen. These exclusions included cases they considered 'sensitive' and children that had been sampled for research by the DWP in the last 3 years.

Table 12.1 Eligible child dates of birth for inclusion in the Growing Up in Scotland study by sample type

12.1.2 Families who moved between sweeps

Any family who moved house between sweep 1 and 2 remained eligible for inclusion in the study as long as they remained in Scotland. The number of sample points covered at sweep 2 was thus slightly larger than at sweep 1 as a small number of families moved outside areas which were initially sampled at sweep 1.

12.2 Response rates

Details of the number of cases issued and achieved and the response rates are detailed in Table 12.2.

Table 12.2 Number of issued and achieved cases and response rates

*-Cases which were considered out-of-scope or unachievable were mostly ineligible addresses - usually due to the family having moved away from Scotland.

12.3 Data collection

12.3.1 Mode of data collection

Interviews were carried out in participants' homes, by trained social survey interviewers using laptop computers (otherwise known as CAPI - Computer Assisted Personal Interviewing). The interview was quantitative and consisted almost entirely of closed questions. There was a brief, self-complete section in the interview in which the respondent, using the laptop, inputed their responses directly into the questionnaire programme.

At sweep 1, primarily because of the inclusion of questions on the mother's pregnancy and birth of the sample child, interviewers were instructed as far as possible to undertake the interview with the child's mother. At sweep 2, interviewers were instructed to undertake the interview with the sweep 1 respondent. Where this was not possible or appropriate, interviews were conducted with the child's main carer. In practice, most interviews were undertaken with the sweep 1 respondent and this was usually the child's mother.

12.3.2 Length of interview

Overall, the average interview lasted around 79 minutes. The child cohort interview had a slightly longer average length at 82 minutes, than the birth cohort interview at 78 minutes. The median interview length for both cohorts was 70 minutes.

12.3.3 Timing of fieldwork

Fieldwork was undertaken over a 14-month period commencing in April 2006. The sample was issued in 12 monthly waves at the beginning of each month and each month's sample was in field for a maximum period of two-and-a-half months. For example, sample 2 was issued at the beginning of May 2006 and remained in field until mid-July 2006.

To ensure that respondents in both samples were interviewed when their children were approximately the same age, each case was assigned a 'target interview date'. For the birth cohort this was identified as the date on which the child turned 22.5 months old, and for the child cohort the date the child turned 46.5 months old. Interviewers were allotted a four-week period based on this date (two weeks either side) in which to secure the interview. In difficult cases, this period was extended up to and including the child's subsequent birthday which allowed a further four weeks.

12.4 Analysis

12.4.1 Weighting

Unlike the sweep 1 weights, a model-based weighting technique was used at sweep 2. All sweep 2 respondents had taken part in the sweep 1 interview; this information on the sweep 2 non-respondents could be used to model their behaviour. Ineligible households (deadwood) were not included in the non-response modelling. ³⁷

Non-response behaviour was modelled using logistic regression. This is a method of analysing the relationship between an outcome variable (in this case response to the sweep 2 interview) using a set of predictor variables. The model takes account of the relationship of the predictor variables to the outcome and the relationships of the predictor variables to each other.

The model generated a predicted probability for each respondent. This is the probability the respondent would take part in the interview, given the characteristics of the respondent and the household collected at sweep 1. Respondents with characteristics associated with non-response (such as being a private tenant) are under-represented in the sample and will receive a low predicted probability. The non-response weights are then generated as the inverse of the predicted probabilities; hence respondents who had a low predicted probability get a larger weight, increasing their representation in the sample.

The final sweep 2 weight is the product of the sweep 2 non-response weight and the sweep 1 interview weight. The final weights were scaled to the responding sweep 2 sample size to give a mean weight of one. This makes the weighted sample size match the unweighted sample size.

12.4.2 Estimating the precision of estimates

Each percentage quoted in this report has an associated margin of error, due to the fact that it is based on only a sample, rather than all children. This margin can be estimated for each proportion, p (where p is the percentage divided by 100) by:

where n is the unweighted sample size (base). This margin corresponds to 95% confidence. In other words there is a 95% chance that the true value across all children in the sub-group (as opposed to just those in the sample) falls within this margin. ³⁸

12.4.3 Presentation of results

Percentages

The percentages shown in the report text have all been rounded to the nearest whole number. Those shown in tables and graphs have been rounded to one decimal place. Consequently, the percentages in one column of a table will not necessarily add to exactly 100.

A dash (-) indicates no respondents at all. All figures are column percentages, except where otherwise indicated.

Bases

Each table shows the weighted and unweighted bases corresponding to each percentage. The data were weighted to compensate for differential non-response across the subgroups. The weighted bases can be used to (approximately) combine two different columns in a table. The unweighted bases can be used to calculate the precision of estimates.

12.4.4 National Statistics Socio-Economic Classification ( NS-SEC)

The National Statistics Socio-economic Classification ( NS-SEC) is a social classification system that attempts to classify groups on the basis of employment relations, based on characteristics such as career prospects, autonomy, mode of payment and period of notice. There are fourteen operational categories representing different groups of occupations (for example higher and lower managerial, higher and lower professional) and a further three 'residual' categories for full-time students, occupations that cannot be classified due to a lack of information or other reasons. The operational categories may be collapsed to form a nine, eight, five or three category system.

The Growing Up in Scotland study generally used the five category system in which respondents are classified as managerial and professional, intermediate, small employers and own account workers, lower supervisory and technical, and semi-routine and routine occupations. Unless otherwise stated, the analysis employs a household level measure of NS-SEC.

Further information on NS-SEC is available from the National Statistics website at: http://www.statistics.gov.uk/methods_quality/ns_sec/cat_subcat_class.asp

12.4.5 Scottish Government Urban Rural Classification

The Scottish Government Urban Rural Classification was first released in 2000 and is consistent with the Government's core definition of rurality which defines settlements of 3,000 or less people to be rural. It also classifies areas as remote based in drive times from settlements of 10,000 or more people. The definitions of urban and rural areas underlying the classification are unchanged.

The classification has been designed to be simple and easy to understand and apply. It distinguishes between urban, rural and remote areas within Scotland and includes the following categories:

Table 12.3 Scottish Government Urban Rural Classification

For further details on the classification see Scottish Executive (2004) Scottish Executive Urban Rural Classification 2003 - 2004. This document is available online at http://www.scotland.gov.uk/Publications/2004/06/19498/38784

12.4.6 Scottish Index of Multiple Deprivation

The Scottish Index of Multiple Deprivation ( SIMD) identifies small area concentrations of multiple deprivation across all of Scotland in a fair way. It allows effective targeting of policies and funding where the aim is to wholly or partly tackle or take account of area concentrations of multiple deprivation.

The first Index ( SIMD 2004) was published in June 2004 and was based on 31 indicators in six individual domains. Data in this report are based on the second index which was published in 2006. The SIMD 2006 contains 37 indicators in seven domains: Current Income, Employment, Health, Education Skills and Training, Geographic Access to Services (including public transport travel times for the first time), Housing and a new Crime Domain.

The SIMD is presented at Data zone level, enabling small pockets of deprivation to be identified. The data zones, which have a median population size of 769, are ranked from most deprived (1) to least deprived (6,505) on the overall SIMD and on each of the individual domains. The result is a comprehensive picture of relative area deprivation across Scotland.

For the purposes of this report, the full index has been separated into quintiles and each case has been assigned a quintile based on the residential postcode. Quintiles are percentiles which divide a distribution into fifths, i.e., the 20th, 40th, 60th, and 80th percentiles. For example, those respondents whose postcode falls into the first quintile are said to live in one of the 20% least deprived areas in Scotland. Those whose postcode falls into the fifth quintile are said to live in one of the 20% most deprived areas in Scotland.

« Previous | Contents | Next »