Internet data collection in the LFS Danish experiences Nordstat, August 2010
Lessons learned from pilot test Outline Lessons learned from pilot test Experiences from full scale CAWI data collection
Background The Danish LFS has individuals as sampling unit Until 2010 selected household variables was collected via proxy in the final interview round Requested by Eurostat to fully comply with regulation no. 577/98 from Q1 2010
Discussions on collection modes: Preparations Discussions on collection modes: How much CATI vs CAWI? How many reminders? The design?
The Pilot Two groups: Aldi vs. Irma 712 household members in each group split over two reference weeks two weeks interview period Aldi: baseline with no reminders and no CATI Irma: Two phone reminders and optional CATI
What was learned: Pilot – the results Reminder needed Indirect effect (threat effect) Direct effect 2nd reminder not necessary A 2 week interviewperiod is sufficient
Number of web-based int. accumulated, reference week 40
Age related effects More lessons CATI option important for older age groups Effect of reminding procedure, relatively larger among younger age groups
Response-rate by age and data-collection method
The Final Setup Important to have time to evaluate pilot thoroughly Going full-scale: Not a big problem 4th panel households selected Only one interview for household members 1 telephone reminder Letters sent Saturday, received Monday
Response rate Q1 2010
Response rate issue – complete households Only complete households are usable 6138 households in Q1 sample 2940 completed interviews with all household members 360 of these households deleted because of reference person
Quality examined - NACE
Quality examined - ISCO
In-depth quality analysis Coding of NACE and ISCO Further work In-depth quality analysis Coding of NACE and ISCO Analysing the mode effects of CAWI in core-LFS
Thank you! Nordstat, August 2010 Michael Frosch mif@dst.dk Sammy Lauritsen ssl@dst.dk