1 Web interviewing at Statistics Netherlands: practices so far Labour Force Survey The Netherlands Björn Janssen Hendrika Lautenbach April 2010
2 Web interviewing : out of control?
3 Why this subject? – Cheap – Efficient – Respondent friendly – High penetration rate of internet
4 Contents of presentation 1. Approach strategy 2. Mode-effects 3. Sample for calculating monthly figures
5 Approach strategy: mixed mode experiences so far CAWI : Letter containing website and log on codes PAPI : Apply for paper questionnaire / return card 1st 2nd 3th 4th ICTCrimeShallow CAWI CAPI / CATI CAWI / PAPI CAWI / PAPI CAWI 1 CAWI / PAPI CAWI CATI (reminder) 2 CAWI / PAPI CAWI CAWI / PAPI CATI method Economy
6 62,7Exp53,041,633,4 Reg51,4 ICTCrimeShallow 68,961,0 Experiments: response levels Crime CAPI: 33% (exp) 58% (reg) Crime CATI: 55% (exp) 71% (reg) Internet: +/- 35 % 67,2 67,6 Economy
7 Approach strategy: deduced from experiences - Whole sample: CAWI -CAWI non response With phone number: CATI Without phone number: CAPI - Response -CAWI: 25-35% -Non response CAWI by CATI: ± 50% -Non response CAWI by CAPI: ± 35% -Total response: ± 60%
8 Approach strategy for subgroups - Plans to develop different approach strategies for selected subgroups of the sample to increase response. Possible subgroups: - Elderly people; - Non-western immigrants; - Large households? - Use of different versions of the introduction letter for the various subgroups.
9 Mode effects on results - CAWI differs from CAPI/CATI: no interviewer present to guide the interview; - The phrasing of the questions as posed in CAPI/CATI will sometimes have to be adjusted to allow independant completion. - Use of different modes has effect on the collected data: unknown which effect.
10 Mode effects: experiences so far - Exchange of respondents - Approaching households not responding to CAWI
11 Sample for calculating monthly figures at this moment -Households which are part of the monthly sample portions are used for calcuting the monthly figures. - The initial monthly sample is used for follow-up every 3 months (panel).
12 Sample for calculating monthly figures when introducing CAWI In the currently chosen approach strategy, the time required to obtain the response of a monthly sample will increase from 1 month to 2-3 months. Which are the options for calculating monthly figures? - the response collected in a calendar month? Which sample should be used for follow-up (panel)?
13 Subjects of discussion Should mode-effects on the data be addressed, and if so, in what way? To increase the response, Statistics Netherlands is considering specific approach strategies for different subgroups. As a result, selected mixed-mode effects for the different groups will be introduced. Is this a problem and should it be avoided? Using a mixed-mode approach including CAWI, how can the time required for data collection be limited to a month which suits the calculation of monthly figures?