Date: 2009-10-05
Specific agreement No 11002.2008.003-2008.211
Technical Implementation Report
for the action Translating and
testing a victimisation survey module
by
The National Council for Crime Prevention, Sweden
1 Introduction
This report from the Swedish National Council for Crime Prevention (Brottsförebyggande rådet) is being written in accordance with Draft Grant Agreement No. 11002.2008.003-2008.211, and relates to the action Translating and testing a victimisation survey module. It constitutes the third of three technical implementation reports to be submitted by the Council in connection with the action, and corresponds to the “quality report” described in Annex IV of the Draft Agreement signed and returned to Eurostat on September 15th 2008.
1.1 Organisation of the report
The report is organised in the following way.
The report begins in Section 2 by recapping the way in which the pilot survey was implemented - sample design, interview mode, timing and interviewer profile. The section summarises the relevant sections of the Technical Report[1] received from the subcontractor (Statistics Sweden) and concludes with a presentation of response rates and a short summary of the subcontractor’s comments on these response rates.
Section 3 of the report presents the views of the survey of Statistics Sweden’s’ interviewers on the basis of questions included at the end of the questionnaire, which they completed at the conclusion of each interview. Section 4 presents an overview of the response frequencies for a number of central variables selected from the different sections of the questionnaire, while Section 5 presents a comparison of data relating to central variables with data collected in an alternative victim survey, namely the Swedish Crime Survey. Section 6 presents conclusions drawn on the basis of the work conducted with the EUCVS pilot, and concludes with a description of the costs involved in the implementation of a full-scale survey in Sweden.
2 Implementation of the pilot survey
As was the case with the expert review and cognitive testing of the victimisation module, the implementation of the pilot survey was subcontracted by the National Council to Statistics Sweden (SCB). Statistics Sweden is a government agency with responsibility for official statistics and other government statistics in Sweden, and has a great deal of experience in conducting high-quality, large-scale statistical surveys of representative samples drawn from the Swedish population in a wide range of areas. Furthermore, the National Council has had and continues to have a successful collaboration with Statistics Sweden in connection with the development and implementation of the Swedish Crime Survey, an annual survey based on a representative sample of 20,000 respondents aged 16 to 79 years, drawn from the general population of Sweden. Within the framework of the constraints associated with the schedule for reporting the results of the Action and the budget adopted for the Action, the National Council has therefore allowed itself to be guided to a great extent by Statistics Sweden in questions relating to what is and is not feasible in the context of a survey of this kind, while at the same time attempting to remain as true to the intent of the Action and the original questionnaire module as possible.
On the basis of experiences gained in connection with the implementation of the Swedish Crime Survey, Computer Assisted Telephone Interviewing (CATI) was chosen as the interview mode. Statistics Sweden were very clear that this is probably not the best alternative for administering the very sensitive section of the module relating to violent victimisation. Other possible methods were considered, but given the financial resources available to the Action, the use of Computer Assisted Personal Interviewing (CAPI) was not an option. Furthermore, the use of this method would make the costs of conducting the full survey prohibitively high. Thus even though certain questions in this section of the questionnaire would no doubt function better in the context of a CAPI interview, it was deemed more realistic from a longer term perspective to test the entire questionnaire using CATI. The particularly sensitive section of the module (referred to hereafter as Section G) was therefore modified for the purposes of adapting it to CATI interviewing.
The study population for the pilot survey comprises individuals who at the end of the year 2009 will be aged between 16 and 79 years. The sampling frame was defined with the help of the Swedish Total Population Register (TPR) and was stratified by county and age groups. A simple random sample (SRS) was drawn from each stratum. The total sample size was 1000 individuals.
The pilot survey was conducted during the period 20th April – 24th May 2009. One week prior to the survey, a cover letter was sent to all individuals in the sample containing information on the contents of the survey and its objectives. They were also informed about relevant confidentiality regulations.
The data collection was carried out by Statistics Sweden’s central telephone group and Statistics Sweden’s field interviewers. The data have been examined in connection with the registration of the interviews. A total of 20 interviewers have participated in the survey.
As was discussed in Report 2, submitted earlier in connection with the Action by the National Council, Statistics Sweden had a number of concerns relating to the suitability of parts of the sensitive Section G for a telephone interview survey. These concerns related to amongst other things the fact that Statistics Sweden’s commissions policy group felt that some of the questions in Section G were so sensitive that they may be experienced as constituting a violation of the respondents’ personal and emotional integrity. The cognitive testing process also showed that both interviewers and those test-persons who had experienced some of the more serious offences included in Section G perceived the interview as sensitive and “difficult”. There were also concerns that the nature of some of the questions was such as to constitute a problem from a work-environment perspective for the interviewers themselves. Statistics Sweden therefore wanted as far as possible to restrict the potential of Section G to lead to respondents experiencing any sense of violation, and to avoid interviewers themselves experiencing excessive negative effects of the interview process, and both these goals were felt to be best achieved by limiting both the number of respondents being asked the questions in Section G in the pilot survey, and the number of interviewers being required to pose these questions.
By restricting the number of respondents being asked the questions from Section G, it was possible for Statistics Sweden to assign this part of the survey to a very small and select group of interviewers. These were hand-picked individuals with substantial experience of sensitive interview surveys. It also allowed resources to be devoted to giving these interviewers special training for this specific survey, it was also possible to provide them with the proper support, inter alia in the form of identified contact persons at the relevant unit of Statistics Sweden, whom they could ring in order to discuss problems or if they were in need of other forms of support.
The sample was therefore randomly divided into two groups. These groups comprised 750 and 250 individuals respectively. For the larger group, almost the entire section G of the questionnaire (which includes questions on amongst other things sexual violence in intimate relationships) was excluded from the interview. The smaller sample group was asked all of the questions in the questionnaire, including those from section G. Response frequencies are presented in Tables 2.1 to 2.3 both for each of the groups separately, and for the sample as a whole.
The total non-response amounts to just over 45 % of the sample, and the majority of the non-response is comprised of individuals who could not be reached. Of the 16.9 % who did not want to participate in the survey, the majority referred to the fact that participation was voluntary. The overcoverage is amongst other things comprised of individuals who have left Sweden.
The technical report sent to the National Council by Statistics Sweden (see Appendix 1) notes that the proportion who were “not-found” in this pilot study is larger than that normally recorded in Statistics Sweden’s interview surveys. If a general comparison is made with the telephone surveys otherwise conducted by Statistics Sweden during the past year, the proportion not-found is approximately ten percentage points higher in the current survey. Statistics Sweden state in their technical report that there are probably a number of reasons for this and that it is difficult to explain what it may be due to with any degree of certainty. Likely reasons may be the nature of the questions (on which information was provided in the cover letter that the respondents received prior to the survey) and the timing of the fieldwork (the length of the data-collection period, and the time of year). A more robust follow-up of the non-response (which would have required more interviewer resources and a longer period of fieldwork) might have been able to generate contact with a larger number of respondents.
Table 2.1 Response rates – Section G excluded
Sample / 750
- overcoverage / 9
741 / 100.0
Responses / 417 / 56.3
Non-response / 324 / 43.7
Of which
unable to participate / 23 / 3.1
not found / 179 / 24.2
declined to participate / 122 / 16.5
Table 2.2 Response rates – Section G included
Category / n / Percent
Sample / 250
- overcoverage / 2
248 / 100.0
Responses / 121 / 48.8
Non-response / 127 / 51.1
of which
unable to participate / 9 / 3.6
not found / 73 / 29.4
declined to participate / 45 / 18.1
Table 2.3 Response rates – Combined
Category / n / Percent
Sample / 1000
- overcoverage / 11
989 / 100.0
Responses / 538 / 54.4
Non-response / 451 / 45.6
of which
unable to participate / 32 / 3.2
not found / 252 / 25.5
declined to participate / 167 / 16.9
The non-response by sex and age-group is shown in tables 2.4a and 2.5a. The corresponding tables for those who participated in the survey by sex and age-group are also shown (Tables 2.4b and 2.5b).
Table 2.4a: Non-response by gender
Gender / n / PercentMale / 249 / 53.9
Female / 213 / 46.1
Total / 462 / 100.0
Table 2.4b: Response group by gender
Gender / n / PercentMale / 264 / 49.9
Female / 274 / 50.1
Total / 538 / 100.0
Table 2.5a Non-response by age-group
Age-group / n / Percent16-29 / 109 / 23.6
30-39 / 94 / 20.3
40-49 / 89 / 19.3
50-59 / 68 / 14.7
60-69 / 64 / 13.9
70-79 / 38 / 8.2
Total / 462 / 100.0
Table 2.5b Response group by age-group
Age-group / n / Percent16-29 / 124 / 23.0
30-39 / 93 / 17.3
40-49 / 86 / 16.0
50-59 / 73 / 13.6
60-69 / 105 / 19.5
70-79 / 57 / 10.6
Total / 538 / 100.0
The tables indicate slight differences between the age and sex distributions of the response and non-response groups respectively, with perhaps the most striking being that the proportion of respondents aged 60 years or over is substantially larger than the size of this group among the non-response. However, Chi-square tests indicate no statistically significant differences between the distributions found among response and non-response groups respectively in relation to either gender or age.
Appendixes 5 through 7 contain the original data files from the pilot survey. Appendix 5 contains the data from the large subsection of the sample, i.e. those who were not asked the questions from Section G. Appendix 6 contains the data from the small subsection of the sample, who were asked all of the questions, including those from Section G. The data from the two subsections of the sample are combined in Appendix 7.
The level of internal non-response to the individual questions was generally very low. In the vast majority of instances, no more than one or two of the respondents asked a given question failed to supply an answer in accordance with the response alternatives provided. Appendix 2 provides a question by question break-down of the responses to each individual item in the survey.
3 Interviewers’ views of the survey
A number of questions posed to the interviewers were included at the end of the questionnaire as a means of gauging their views on amongst other things the clarity of the questions and the respondents’ comprehension of the questions and willingness to answer. The responses to these questions are presented in Appendix 3.
In summary, the answers to these questions indicate that the interviews worked at least reasonably well. A total of over 90 percent of respondents “never” (61.0%) or “almost never” (29.7%) had to ask for a question to be clarified, and the respondents were “never” (89.9%) or “almost never” (7.1%) unwilling to answer any questions in over 95 percent of the interviews. The vast majority of respondents were perceived as having tried to answer the questions as well as they could “very often” (87.4%), with a further eleven percent being perceived as having done so “often”. The interviewers felt that the respondents for the most part understood the questions “often” (24.3%) or “very often” (73.2%) in over 97 percent of the interviews and in over 90 percent of the interviews, the interviewers had no concerns that the answers to one or more sections of the questionnaire were of lower quality than the answers provided to the other sections (e.g. as a result of tiredness, concentration difficulties or the nature of the questions).
The final question in this section of the questionnaire directed at the interviewers asked whether they had any additional comments on the interview. In over 90 percent of the cases, the interviewers felt they had nothing more to add. Their open answers in the remaining cases (which have been translated in full in Appendix 4), indicate for the most part some concerns/suggestions relating to: a) specifying the “household” referred to in the questions, particularly among younger respondents who shift residence quite often, since a period of five years may in their case have involved membership of a number of different households, b) there were a number of suggestions for improved wording/better translation in some of the items, c) a number of requests for further clarification of definitions in relation to various questions, d) a couple of cases where the interviewer felt it would be more logical to have the question on whether the respondent possessed a car prior to the question on whether he/she worried about having it stolen or damaged.
4 Overview of response frequencies
This section of the report presents the response frequencies for a number of central variables included in the questionnaire both for the sample as a whole, and in the majority of cases also by gender and age. It begins with variables relating to the section of the questionnaire focusing on feeling safe and the respondents’ worries about crime and then moves on to present response frequencies relating to first household victimisation and then other forms of victimisation.
4.1 Section B – Feeling safe and worries about crime
Table 4.1 presents response frequencies for the question: “How safe do you feel walking alone in the area where you live after dark?” (Question B3). As can be seen from the table, the majority of respondents reported feeling either very or fairly safe when walking alone in the area where they live after dark. At the same time there are differences among the respondents by both age and gender as regards both the proportion reporting that they don’t go out at night, and that they feel safe walking alone in the area where they live after dark. Females and the oldest group of respondents are more likely than males and the younger groups of respondents respectively to report that they do not go out after dark.
Table 4.1 Response frequencies for question B3. (How safe do you feel walking alone in the area where you live after dark). All respondents and by sex and age. Percent.
Response frequencies (percent)Very safe / Fairly safe / A bit/very unsafe / Never goes out
All respondents / 50.8 / 30.6 / 6.8 / 11.8
Males / 67.4 / 21.8 / 3.8 / 6.9
Females / 34.9 / 39.0 / 9.6 / 16.5
16-25 years / 57.9 / 25.3 / 9.5 / 7.4
26-35 years / 54.8 / 33.3 / 7.1 / 4.8
36-45 years / 58.7 / 27.2 / 5.4 / 8.7
46-55 years / 50.0 / 39.2 / 5.4 / 5.4
56-65 years / 45.8 / 34.4 / 7.3 / 12.5
66-79 years / 38.0 / 26.1 / 5.4 / 30.4
Of those who do go out after dark, 92.3 percent of respondents reported feeling very/fairly safe, and 7.7 percent feeling a bit/very unsafe. In relation to gender, (once again, among those who actually go out after dark) the proportion reporting feeling a bit/very unsafe ranges from 4.1 percent among the male respondents to 11.5 percent among their female counterparts. In relation to age, there is something of a U-curve, with the largest proportions reporting feeling a bit/very unsafe being found among the 16-25 year-olds (10.2%) and the 56-65 year olds and 66-79 year olds (8.3% and 7.8% respectively), and the smallest proportions being found among those aged 36-45 (6.0%) and 46-55 (5.7%); figures once again based on those who do actually go out after dark at least occasionally.
The corresponding information relating to the responses to question B6 (“How safe do you feel in your home at night?”) is presented in Table 4.2. The table shows on the one hand that very few respondents report feeling even a bit unsafe in their homes at night (no respondents reported feeling “very unsafe” in their homes at night), and on the other that the variations in relation to age and gender are less prominent than they were in relation to the previous question. If anything, males do appear slightly more likely to report feeling very safe than their female counterparts. And in relation to age, a similar U-curve can be noted in the proportions reporting feeling a bit unsafe – although the numbers of respondents involved here (given the size of the sample and of the proportion reporting feeling even a bit unsafe in their homes at night) are very small.
Table 4.2 Response frequencies for question B6 (How safe do you feel in your home at night?). All respondents and by sex and age. Percent.
Response frequencies (percent)Very safe / Fairly safe / A bit unsafe
All respondents / 81.6 / 16.7 / 1.5
Males / 85.2 / 12.9 / 1.9
Females / 78.5 / 20.4 / 1.1
16-25 years / 83.2 / 14.7 / 2.1
26-35 years / 81.2 / 17.6 / 1.2
36-45 years / 89.2 / 10.8 / 0
46-55 years / 82.4 / 17.6 / 0
56-65 years / 76.3 / 21.6 / 2.1
66-79 years / 78.5 / 18.3 / 3.2
Table 4.3 presents response proportions for the questions on how worried the respondents had been about exposure to specific types of crime in the form of burglary, theft/damage to their car or being attacked by a stranger (Questions B8, B10 and B12).