260 likes | 270 Views
Analysis of data collection methods, respondent selection, interview length, recall periods, and data transmission timing. Recommendations for improving quality and cost-efficiency in survey practices.
E N D
EVA group asked to consider five issues Mode of data collection Selection of respondents Length of interview Recall period Time limit for data transmission Age limit for respondents
1 – Mode of interview – the pilots Variety of modes across country; several countries tried more than one mode
1 – Mode of interview No conclusive pattern for response rates But PAPI & CAPI rates a bit higher than CATI CAWI rate in Finland low Implications for costs PAPI likely to be most expensive (data processing) CAPI costs roughly double that of CATI CAWI might be cheapest
1 – Mode of interview Implications for quality CATI & CAPI rates much on a par? PAPI quality poorer CAWI – largely unknown Future developments Traditional CATI will face problems (mobile phones) BUT industry will have to respond CAWI likely to be heavily developed, and could be offered as an option
1 – Mode of interview RECOMMENDATIONS Interview mode as standardised as possible CATI seems best option (fair degree of consensus) Though CATI may not be possible for some countries And mobile phones need to be included Further tests of CAWI should be taken on board
2 – Selection of respondents ISSUES One respondent per household? Or ‘most responsible’ for household crimes Bears on lower age limit Substitution……….
3 – Length of interview ISSUES Lengthier interviews cost more And can lead to high refusals When told average interview length And when get ‘fed-up’ Experience shows 20 minutues is maximum for CATI Likely to be the same for CAWI
2 – Selection of respondents One randomly selected respondent most cost efficient can be from national register of persons, or random person selected from a household register Young respondents might not be best placed to answer about household crime. BUT Much more expensive to interview more than one person per household And would reduce response rates
2 – Selection of respondents RECOMMENDATIONS One person per household Who are asked about both household and personal crimes No substitution Well-established and cost efficient approach
3 – Length of interview Current EVA Group questionnaire much shorter than existing questionnaire – because Section G omitted in current form Reduced number of questions – e.g. Fear of crime Security precautions Personal and household characteristics Details of victimisation incidents
3 – Length of interview But may be still too long? Reducing average questionnaire length would Cut costs for a given sample size Give bigger samples for a given cost Allow room for ‘rotating’ or ‘one-off’ modules
3 – Length of interview OPTIONS Reducing ‘ask-all’ questions – e.g. Attitudinal / perceptions questions Personal and household details Asking attitudinal / perceptions questions of sub-samples 1 and 2 reduces information and complicates analysis
3 – Length of interview MORE RADICAL THIRD OPTION Core 2,000 (?) EU-SS sample – all questions Allowing basic cross-sectional analysis Plus extended sample for victimisation experience only Reduction of 30% (?) on length of core questionnaire Better estimates of victimisation Sub-national estimates BUT higher costs
3 – Length of interview RECOMMENDATIONS While current EVA Group questionnaire is nearly acceptable, consideration should be given to reducing it more Further thought needs to be given to: Cost savings of Options 1 & 2 Information loss of Options 1 & 2 Cost and methodolgical implications of Option 3
4 – Recall period and timing of fieldwork ISSUES Memory loss = shorter recall period But risks ‘forward telescoping’ (more serious incidents) Also need enough victimisation incidents = longer recall period Calendar year or ‘last 12 months’ recall period Calendar year more understandable But needs fieldwork early in the year
4 – Recall period and timing of fieldwork CURRENT EVA GROUP QUESTIONNAIRE Standardised recall period Last five years / last (calendar) year & how often / this year Victimisation details for ‘last incident’ over five years Last (calendar) year gives 1-year prevalence and incidence victimisation estimates
4 – Recall period and timing of interview RECOMMENDATIONS Last five years / last (calendar) year / this year best option Shown to have worked in the past Needs fieldwork early in the year If this is not generally feasible, fieldwork timing should be standardised And further consideration given to a calendar year or ‘last 12 months’ recall period
5 – Time limit for data transmission ISSUES EU-SS results need to be timely Data needs to be processed quickly But EVA group unsure about current Eurostat data transmission requirements Weighting for different modes…….
5 – Time limit for data transmission RECOMMENDATIONS Investigate current Eurostat data transmission requirements
6 – Age limit for respondents THE PILOTS Little consistency Most pilots 16+ with no upper age limit Some 18+ only A few age 15+ Most no upper age limit, but some had a limit And for Section G
6 – Age limit for respondents ISSUES – YOUNGER AGE LIMIT Young respondents may not respond well on household victimisation But there are ‘important victims’ And capturing them for personal but not household crime is expensive (see above) What is young – 15, 16, 17, 18? Parental permission
6 – Age limit for respondents ISSUES – OLDER AGE LIMIT Older respondents Cognitive ability Sensitivity as regards personal violence questions Limit for everything or just violence & sexual victimisation or just sexual victimisation
6 – Age limit for respondents RECOMMENDATIONS Age range should be standardised Starting at age 16 would maintain consistency with other surveys BUT 18 might be more workable EVA Group see no strong argument for setting an upper age limit BUT majority view seems to favour a limit 79?