covering the general population by internet interviewing marcel das 19-9-2015 1
TRANSCRIPT
Covering the general population by Internet
interviewing
Marcel Das
19-04-23
1
MESS Project
• core element: LISS panel
• other key elements:
- core study
- experimental modules
- new forms of data collection
- links with administrative data
- special groups
19-04-23
2
Core element: LISS panel
Online panel of 5,000 households 8,000 individuals (>= 16 years)
Questionnaires each month, 30 min.
Incentive 15 euro an hour (average)
For scientific use, at no cost
19-04-23
3
LISS panel Online interviews as method, but:
Probability sample drawn from address sampling frame of Statistics Netherlands
Includes households without Internet access (less than 15%): CentERdata provides equipment
Contacted by letter, telephone or visit19-04-23
4
simPC
Very small and silent
Only the most frequently used functions
Automatic maintenance, safety
Simple operation and readable screens
Installation and support19-04-23
5
19-04-23
6
Internet interviews
letter and brochure
short interview
simPC and adsl
households with Internet
households without Internet
confirmation
panel question
Recruitment
19-04-23
7
Internet interviews
letter and brochure
short interview
simPC and adsl
confirmation
panel question
Primary response
target: 80%
Secondary response
target: 60%
Tertiary response
target: 53%
Response
Recruitment
experiment
19-04-23
8
Experimental design CATI CAPI N
Special letter
Standard letter
Special letter
Standard letter
Prepaid incentive
Introduce panel in letter 0 euro 10 euro 200
Introduce panel after interview
20 euro 50 euro 300
Promised incentive
Introduce panel in letter
0 euro 10 euro 400
Introduce panel after interview
50 euro 20 euro 300
N 200 200 400 400 1200
CATI = Computer Assisted Telephone InterviewingCAPI = Computer Assisted Personal Interviewing
19-04-23
9
Pilot experimental groups1. CATI: special letter - no incentive - Panel in letter and
brochure
2. CATI: special letter - promised incentive E. 50 - Panel introduced after interview
3. CATI: Standard letter - prepaid incentive E. 10 - Panel in letter and brochure
4. CATI: Standard letter - prepaid incentive E. 20 – Panel introduced after interview
5. CAPI: special letter - prepaid incentive E. 50 – Panel introduced after interview
6. CAPI: special letter - no incentive - Panel in letter and brochure
7. CAPI: Standard letter - promised incentive E. 10 - Panel in letter and brochure
8. CAPI: Standard letter - promised incentive E. 20 - Panel introduced after interview
19-04-23
10
Pilot natural groups
CAPI
1. Unknown or no telephone number: 30%
CATI and CAPI
2. Known (landline) telephone number: 70%
19-04-23
11
Observed response in CATI versus CAPI by subpopulation CATI
households with telnr
%
CAPI Households with telnr
%
CAPI Households
without telnr %
Primary response 65 58 44 Primary refusals 22 25 25 Primary not reached 9 14 25 Primary not usable 3 3 5
Effect of mode (1)
19-04-23
12
Observed response in CATI versus CAPI by subpopulation CATI
households with telnr
%
CAPI Households with telnr
%
CAPI Households
without telnr %
Primary response 65 58 44 Secondary response conditional on primary
67 84 88
Primary * secondary 45 50 39 Tertiary response conditional on secondary
80 72 64
Primary * tertiary 35 35 25
Effect of mode (2)
19-04-23
13
incentives increase response rates
- effect is large from 0 to 10 euro (increase of 15%)
- effect is not significant from 10 to 20, 20 to 50 euro (about 3%)
prepaid works better than promised
Effect of incentives
19-04-23
14
Effect of information
no effect of letter content
no effect of timing panel introduction
effect of seeing/reading brochure
Set up main recruitment
• Combination CATI – CAPI, follow up CAPI
• Prepaid 10 euro incentive
• Promised extra 10 euro incentive for starting
• Attention to design letter and brochure
19-04-23
15
Results
19-04-23
16
Pilot versus main response rates
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
primary secondary tertiary
response
pro
po
rtio
n o
f to
tal
Pilot
Main
Non-response patternsSimilar to those of other leading
scientific panels
Superior to commercial access and volunteer panels:
- no coverage problems
- no selfselection
19-04-23
17
Attention for difficult groups
Correction by oversampling in refreshment of 2009 (with Statistics Netherlands)
New immigrant panel (with Cross-cultural Psychology, Tilburg University and Statistics Netherlands)
19-04-23
18
Correction of bias by stratified refreshment
sample 2009
19-04-23
19
GroupPopulation
2007 %
April 2008Bias
April 2009Bias
Jan 2010 Bias
Age >=70 11.6 -5.4 -5.4 -2.8
Living alone 19.2 -5.4 -5.0 -2.4
Non-western immigrant
9.6 -5.1 -4.8 -3.7
Persons
New forms of data collection
19-04-23
20
Use of graphical or animated presentation
19-04-23
21
Biomarkerspilot study: blood cholesterol, saliva cortisol, and waist
circumference
19-04-23
22
More biomarker experiments...- accelerometers
- advanced scales (based on bioelectrical impedance)
link objective data with self-report data
high frequency measurement
19-04-23
23
Use of the LISS panel
19-04-23
24
Questionnaires
The “household box”
Experimental modules open to academic researchers at no cost
Core study
19-04-23
25
Experimental modules
Who can use the LISS panel:every researcher who wants to collect data for scientific, policy or societal relevant research
Scientific research: Academic researchers, irrespective of nationality, can use the LISS panel at no cost
Proposals can be submitted throughout the year
19-04-23
26
19-04-23
27
Examples: subsample
Effects of illness babies in first year on labour participation parents:
Pregnant women, follow 3 years!
Adolescent/teen identity and the Internet:
12-18 year olds
Random subsample or experimental groups
19-04-23
28
Procedure:
Submit to: [email protected]
Typically be between 2 and 5 pages
Evaluation: Board of Overseers
Decision: 1-3 months Accept Revise and resubmit Reject
(http://www.lissdata.nl/lissdata/Proposals/How_to_Submit_a_Proposal)
19-04-23
29
Criteria:
Scientific potential: predicated on literature; sound analytical
framework
Fit in LISS panel: value of using an Internet survey
Feasibility: questions doable for respondents?
Burden: length and target sample
(http://www.lissdata.nl/lissdata/Proposals/How_to_Submit_a_Proposal)
19-04-23
30
Requirements:
Data protection statement
Single user access
Publication list
Data available to other researchers!
Proposals: Summary (May 2010)
19-04-23
31
Decision Board of Overseers Count
Undecided (in review or revision) 14
Accepted 49
Rejected 10
Total 73
Proposals: Examples
19-04-23
32
Proposal Affiliation(s) Perceptions of nutrigenomics Twente University (NL)
Mental health Radboud University, Nijmegen (NL); Emory University (USA)
Subjective well-being Max Planck Institute for Human Development, Berlin (Germany)
Computer crime victimization Leiden University (NL)
Volunteer work University of Duisburg / Mannheim (Germany)
Open-ended questions University of Michigan (US)
Identification of guilt sensitivity University of Laval (Canada)/Copenhagen (Denmark)
Available data Who can use the LISS data:every researcher who wants to use data for scientific, policy or societal relevant research
Use of data is free of charge (unless data are used for research funded by means of external sources)
Available data: longitudinal core study proposed studies
19-04-23
33
Longitudinal core study
Household and familyEconomic situation and housingWork and schoolingSocial integration and leisureHealthPersonalityReligion and ethnicityPolitics and values
19-04-23
34
LISS data
Disseminated through website / data archive:
http://www.lissdata.nl
Database is based on an internationally used specification for describing social science data (DDI)
One of the first implementations of DDI version 3
19-04-23
35
8-6-2010
36
Why use DDI? A lot more work?
Simple links to datasets and codebooks: fine for single datasets and once-a-year panel studies
LISS data: complex and extensive
Will the data and documentation still be accessible in 10 years?
8-6-2010
37
Complex data
Q5Panel
members
Q8Q7
Q6
Q1 Q2 Q3 Q4
Core questionnaire
ProposalsQ9
8-6-2010
38
Extensive data Example: “Do you know which panel
members are vegetarian?” Asked somewhere, but in which study?
Example: “Did you ever measure political efficacy?”
Asked sometime, but when exactly?
Example: “Can I estimate the housing expenditure of impulsive persons?”
Which items measured impulsiveness?
8-6-2010
39
DDI 3 & LISS website
Search entry with keyword and on concept, topic, and study (relational)
Can handle longitudinal data (waves)
Exchangeable with other data archives
Flexibility for future applications (Example: iPod version)
October 2010 Publisher: Routledge / Taylor & Francis Group
19-04-23
48