320 likes | 438 Views
18 July 2013. ESS Mixed mode experiment. ESRA Conference, Ljubljana, July 2013 Alun Humphrey @alunhumphrey. Contents. Background and experimental design Response rates Sample profile Completion online Costs. Background. 1. Funders. Co-funded by
E N D
18 July 2013 ESS Mixed mode experiment ESRA Conference, Ljubljana, July 2013 Alun Humphrey @alunhumphrey
Contents Background and experimental design Response rates Sample profile Completion online Costs
Background • 1.
Funders Co-funded by • Economic & Social Research Council • City University, London
Key aims Feasibility of online ESS • UK only has address-based sample frame • No personal contact at address – letter only • Need to carry out random selection Cost effectiveness of online ESS • Need for two survey instruments (F2F/ web) • How best to incentivise web completion • Are there cost savings?
Key issues Exploring online response and subsequent F2F response rates Assessing the impact of different value incentives Identifying best approach to carrying out random selection in absence of interviewer Exploring how people complete long interviews online
The sequential design Online questionnaire Face-to-face follow-up
Design Run alongside main European Social Survey Round 6 Sample of 2,000 addresses from Post Office Address File All sent letter plus £5 incentive Postal and email reminders if no response Key experimental elements • Two methods of respondent selection tested (birthday vs web selection) • Two conditional incentive amounts (£15 vs. £35) • 500 addresses in 20 points then visited by interviewer who attempts F2F interview with £5 conditional incentive
Contact strategy Pre-notification Letter 7 days Mailed invite to web survey with £5 incentive 7 days Reminder letter 10 days Interviewer visit 2 months
Response • 2.
Response rates • Base: All eligible addresses
Impact of incentive • Base: All eligible addresses
Selection method • Base: All eligible addresses
Sample profile • 3.
Sample profile - gender Female Male • Base: All respondents
Sample profile - age 15-24 25-44 45-64 65+ • Base: All respondents
Sample profile – years in education 16-19 20 or more Less than 9 9-12 13-15 • Base: All respondents
Sample profile – in work • Base: All respondents
Sample profile – household income Mid Low High Refusal DK NA • Base: All respondents
Completion online • 4.
Web survey completion levels • Base: All web respondents
81% of responders completed the web questionnaire in one session. (80% fully completed and 1% partially) • Base: All web respondents.
Length of web sessions Median sessions length was 42 minutes But some took extra sessions to complete web survey so median total length was 52 minutes – very similar to F2F ESS R6
Does random selection by respondents work? 68% comply Of eligible respondents who provided information on household Only 40% of eligible provided information on household. We don’t know about the other 60%. • Eligible: Web respondents who reported more than 1 adult living in the household.
Costs • 5.
Conclusions Higher incentive unnecessary Potential problems with person selection done by household Some profile differences by mode, but much reduced in combined web plus face-to-face sample Potential for cost savings using combined approach
Thank you If you want further information or would like to contact the author, • Alun Humphrey • Senior Research Director • T. 020 7549 7043 • E. Alun.Humphrey@natcen.ac.uk • Visit us online, natcen.ac.uk • Follow us on @alunhumphrey • @natcen