British Crime Survey
Options for extending the coverage to children and people living in
This report presents findings from methodological research carried out to examine the feasibility of covering children (under 16s) and people living in communal establishments as part of the BCS or as separate surveys. It reviews highlighted criticism about the coverage of the BCS, which is currently restricted to measuring crimes experienced by adults resident in private households and excludes crimes committed against: under 16s; those living in institutions, communal establishments or on the streets; and businesses.
Title: British Crime Survey: Options for extending the coverage to children and people living in
Authors: Kevin Pickering, Patten Smith, Caroline Bryson and Christine Farmer
Series: Home Office Research Report 06
Number of pages: 22
Date published: May 2008
Availability: Download full report (more download options )
The main aims of the methodological work were:
- to examine the feasibility of covering children and those living in communal establishments as part of the main BCS or as separate surveys
- to outline different options for obtaining nationally representative estimates of crimes against these groups, as well as estimates of crime-related perceptions
Children
One of the aims of the research was to assess the approaches that could be taken to obtain a sample of children and to investigate features of the design, in particular: how to obtain the sample; the appropriate age range; the survey mode; ethical considerations and sample size.
Children aged younger than 10 years should not be included in the BCS child sample, as they are less likely to have the appropriate cognitive skills and the non-coverage of this group is likely to have only a marginal impact on victimisation estimates for the under 16s. Restricting the survey to those aged 10-15 means that a single questionnaire could be produced to cover the whole age range, thus reducing complexity and costs. This would not be the case if children younger than 10 were included. The information collected from the child sample would need to be analysed separately from the main BCS data, so the questions could be written to ensure that they could be completed by children aged 10-15.
The best approach to obtain a sample of under 16s would be to select them from the households that were selected for the main BCS survey rather than from alternative sources such as child beneft (CB) records or via schools. Children in the eligible age range (10-15) would be identifed in 15% of households in the main BCS sample. By selecting children in these households, adequate sample sizes to obtain nationally representative estimates by age group and sex would be achieved.
The children should either be interviewed using computer- assisted interviewing (CAI) a mix of computer-assisted personal interviewing (CAPI), computer-assisted self interviewing (CASI) and audio-CASI or the children should complete a paper questionnaire. The former would produce higher quality data with great fexibility for the interview, but at a higher cost. The authors recommend that only one child should be interviewed in the household if CAPI, CASI or audio-CASI is used, so as to minimise the burden on households. If a paper questionnaire is used, then more than one child in a household could be included. This choice does, therefore, also impact on sample sizes. With a paper questionnaire it would be possible to achieve a sample size of over 6,000 children. With the computer interviewing, the sample size would be about 5,500.
Communal establishments
To date, BCS has not covered the population of communal establishment residents because it comprises a small proportion of the adult population and would be costly to cover. This research examined whether this conclusion remains justifed for the BCS. The communal establishment resident population made up about 2.1% of the adult population at the time of the 2001 Census; the BCS already covers nearly 98% of the population aged 16 and over. Extending the BCS to cover communal establishments would, therefore, have little impact on the overall estimates of the prevalence of victimisation produced from the BCS.
No lists of people living in communal establishments currently exist and any sampling method would therefore require identifcation of residents at a previously drawn sample of institutions. The authors identify two feasible approaches to sampling communal establishments:
The first is to screen a Postcode Address File (PAF) sample for the presence of institutions. Small-user PAF addresses would be screened during main BCS feldwork. A larger-user PAF sample would be drawn and screened using the Inter-Departmental Business Register (IDBR). This approach proved to be workable in a pilot study by the Offce for National Statistics (ONS). However, because institutions vary widely in size, this approach will only give statistically effcient estimates if large institutions are sampled at higher rates than small institutions. It may be possible to achieve this to a limited extent by adjusting the relative small-and large-user PAF address selection probabilities, although recent changes to the structure of PAF may render the method ineffective. Efficient samples of residents could be drawn if size information were available for a large initial sample of institutions, as this would enable a sample of institutions that was disproportionately stratifed by size to be drawn. The initial institution sample would need to be very large, and because each member of the initial sample would need to be contacted for size information, the cost of this approach would be considerable.
The second approach would be to compile a frame of communal establishments from available listings. This would require considerable work, and would probably not result in a fully comprehensive frame. It may prove feasible to construct comprehensive frames of residential and care homes, educational establishments and nurses’ homes for BCS use, although this would also require considerable time and effort. These would cover about two-thirds of the institutional population.
In conclusion, it would not be feasible to cover communal establishment residents in BCS in a statistically reliable manner without incurring very substantial additional costs. If a decision were made to cover this population, a properly resourced feasibility study would be essential.
Getting a copy
Last update: Tuesday, May 20, 2008