« Previous | Contents |
Listen
Footnotes
1. McCaig, E and Leven, T (2003) Fundamental review of the Scottish Crime Survey, Edinburgh, Scottish Executive. In the remainder of the text this will be referred to as 'the McCaig Review' or simply 'the Review'.
2. Op cit p.52.
3. This was calculated by excluding all non-contacts so was unrealistically reassuring about the response rate penalty of switching to a telephone survey. This response rate is not comparable with the response rates reported below for either the telephone or face-to-face surveys.
4. It is not clearly spelled out in the Review what is meant by "poorer quality information". We have taken it to mean higher item non-response than would be expected in a face-to-face survey. Similarly it is not made clear why a telephone survey would have problems with sensitive topics. We have interpreted this as referring to the difficulty of the interviewer establishing rapport with respondents on the telephone and so people being reluctant to discuss sensitive issues such as domestic violence or sexual assault.
5. op cit, p.60 although the Review does not make clear what level of variation would be required to be "worrying".
6. The Scottish Household Survey is a continuous general household survey of the Scottish population in private residences. Over a two-year sweep over 30,000 face-to-face interviews are conducted across Scotland in respondents' homes. The sampling is designed to be representative of the Scottish population each quarter, to permit analysis of the larger local authorities and groupings of smaller authorities every year and to allow analysis at a local authority level every two years. The survey is conducted by MORI Scotland and TNS Social using computer-assisted personal interviewing and achieves a response rate that is typically 67-68%. More details can be found at www.scotland.gov.uk/shs . At the time of undertaking the analysis, data from 2003 was the most recent publicly available.
7. The Scottish Household Survey is also carried out by MORI Scotland and TNS Social. At the time of sampling for 2003 and 2004, 2002 was the most recent year for which sample outcome information was available.
8. The short interviews comprised only basic demographic information and a self-completion component to collect information on drugs misuse and domestic violence. Neither of these is relevant for comparison with the telephone survey and the short interviews are not discussed further. Version A and Version B indicated different follow-up sections of the questionnaire which came after the incident screening and Victim Forms.
9. For example, special enumeration districts (military bases, prisons, large communal establishments etc) are excluded from samples drawn from PAF so a private residence in a special ED will be excluded from a PAF sample but might be included in an RDD sample.
10. For example, Oldendick, Robert W. (1988) "A Comparison of the Kish and Last Birthday Methods of Respondent Selection in Telephone Surveys" in Journal of Official Statistics, Vol.4, No.4, 1988. pp. 307-318. More recently Tipping, S and Nicolaas, G (2001) "Respondent Selection procedures for Telephone Surveys" in Survey Methods Newsletter, Vol. 21, No.1.
11. Gaziano, C. (2005) "Comparative Analysis of Within-Household Respondent Selection Techniques" in Public Opinion Quarterly 69(1):124-157.
12. cited by anonymous reviewer as O'Rourke D and Blair J (1983) "Improving random selection in telephone surveys", Journal of Official Statistics 10: 428-32.
13. Or more accurately a probability of selection determined only by design features of the survey such as disproportionate stratification.
14. Although it is difficult to say to what extent the focus of the SCS on private households is simply a reflection of the sampling frames that have been available for the survey. Clearly, the omission of significant groups of the population (such as students in halls or prisoners) who might experience higher levels of victimisation would be a problem if the objective were to measure all forms of victimisation.
15. Most notably for this study, the International Crime and Victimisation Survey ( ICVS). Other significant studies, such as Ashworth, Wands, Turtle, Lloyd (2000) Police Public Consultation: Developing a Model Survey. Home Office, London have followed the practice used on the ICVS of treating all non-contacts treated as ineligible.
16. In Ashworth et al these lines are discounted as calls boxes, second homes and businesses. In the SCVS, daytime interviewing would have identified and screened out business and as we show in Appendix 1 call boxes and second homes cannot possibly account for the level of non-contact.
17. for example, Nicolaas, G and Lynn, P (2002) "Random-digit dialling in the UK: viability revisited" in Journal of the Royal Statistical Society A Vol.165, Part 2, pp.297-316.
18. See www.aapor.org . There is no guidance on how the estimate might be arrived at although it is clear that a genuine attempt to make an informed estimate should be made rather than arbitrarily picking a number.
19. McCaig and Leven, op cit, noted that "while this [non-coverage of mobile-only households] may be an issue for the future, the proportion of homes without a landline phone remains small" p.54.
20. Taylor S (2003) "Telephone surveying for household social surveys" in Survey Methodology Bulletin No.52 July 2003, Office for National Statistics.
21. Ofcom (2004) Consumers' use of fixed telephony August 2004.
22. The first wave of MORI Scotland's Social Policy Monitor, with a sample of 1,080 households asked the same questions as were used in the face-to-face survey.
23. These survey estimates are subject to some sampling error - the face-to-face survey estimate suggests that the true value in the population as a whole is likely to be in the range 6.1% to 7.9% while the later MORI Scotland estimate suggests the true value is likely to be in the range of 7.3% to 10.7%.
24. UK Changes who provide samples to both MORI and BMRB.
25. McCaig and Leven, op cit, p.57.
26. This analysis was undertaken with data weighted only to account for design elements. The telephone survey was weighted to account for disproportionate sampling by Police Force Area and to account for some households having more than one landline for receiving calls. Respondents were also weighted by the inverse of the number of adults in the household to account for the fact that as the number of adults in a household increases, each adult has a correspondingly reduced probability of selection. The face-to-face survey was only weighted to account for the number of adults in households. The Scottish Household Survey has no corrective weights applied. This does not mean that the SHS is unbiased but it is the closest comparator available and represents what might be achieved from a face-to-face crime survey of a similar size and design.
27. Rizzo, L. Brick, JM and Park, I (2004) "A minimally instrusive method of for sampling persons in random digit dial surveys" in Public Opinion Quarterly 68:2.
28. See, for example, McVie et al (2004) Scottish Crime Survey 2003 Scottish Executive Social Research
29. It is possible that this is incorrect. An error in the telephone survey script meant that no information was collected about how many of the incidents reported in a series occurred within the survey reference period. Whether any incidents occurred in the reference period was known and in most cases the number of periods in which incidents occurred equalled the total number of incidents. The incidents could therefore be accurately assigned to periods. In around 20% of cases it was necessary to estimate the number that might have happened in the reference period by assigning incidents evenly across all the periods in which incidents were known to have occurred. For example, if 5 incidents occurred in two periods, each period was given 2.5 incidents.
30. Both companies used the same coding manual and a 5% sample of each company's coding was submitted to the surveys' project manager at the Scottish Executive for independent verification.
31. For example, research by Ofcom shows that 21% of adults use a mobile as their main way of making and receiving calls. (Ofcom (2005), The telecommunications market 2005 chart 3.46 p.135).
32. The American Association for Public Opinion Research. 2004. Standard Definitions: Final Dispositions of Case Codes and Outcome Rates for Surveys. 3rd edition. Lenexa, Kansas: AAPOR.
33. Calculated as the percentage reduction in the mean square error between the reference data and the unweighted data and the reference data and the weighted data.
« Previous | Contents |