Measuring Interviewer Effects on Survey Error in SHARE

Download Report

Transcript Measuring Interviewer Effects on Survey Error in SHARE

Measuring Interviewer
Effects on Survey Error in
SHARE
Annelies Blom
Julie Korbmacher
Ulrich Krieger
Motivation
 Korbmacher and Schröder (2010): consent to record
linkage (SHARE wave 3)
„The decision making process is mainly influenced by the
interview situation which in turn is driven by the
interviewer-respondent-interaction ”
 …make contact
 …gain cooperation
 …ask survey questions
 …conduct measurements
The and
rolemeasurements
of the interviewer
 …record answers
 …maintain respondents’ motivation throughout the
interview
Standardized interviews to
reduce variation in the
entire data collection
process
Types of interviewer effects in surveys
Unit nonresponse
Contact
Interviewer
Cooperation
NonresponseError
Item nonresponse
MeasurementError
Measurement
Survey of Health, Ageing and Retirement
in Europe (SHARE)
 Multidisciplinary
 Micro data on health, socio-economic status, social and
family networks
 bi-annual, longitudinal (wave 1 in 2004)
 20 European countries
 SHARE-DE (Wave 4) specials:
Record linkage (Pilot in wave 3)
Collecting biomarkers
Nonresponse experiment
Interviewer effects examined in SHARE
Germany Wave 4
Consent to biomarkers
Consent to record linkage
blood pressure
height
waist circumference
dried blood spots
Unit nonresponse
(incentives)
Interviewer
No unconditional incentive
10€ unconditional incentive
20€ unconditional incentive
40€ unconditional incentive
Income item nonresponse
Measuring and Explaining interviewer
effects
 Step 1: Measuring interviewer effects
 Step 2: Explaining interviewer effects
 Who are the SHARE interviewers?
Interviewer questionnaire
Underlying assumptions:
Interviewers differentially impact on the data collection process
This differential impact is related to their – conscious and
subconscious – appearance and actions
These actions can be explained by characteristics collected in
an interviewer survey
Interviewer survey
 2011
 Interviewer Training Wave 4 (trained 197 interviewers)
 Paper-and-pencil
 Voluntary and no incentives
 At the end of the training session
 Response rate: 83%
 Link via InterviewerID to
SHARE Survey data!
Conceptual framework
 4 dimensions of interviewer
characteristics
 5 aspects of SHARE Wave 4 (Germany)
Unit nonresponse
General
attitudes
Own behavior
Experience with
measurements
Expectations
Unit nonresponse
(incentives)
Consent to
biomarker
collection
Consent to
record
linkage
Item
nonresponse
(income)
Conceptual framework
 General interviewer attitudes
 Reasons for being an interviewer
 Attitudes towards best practice
 Trust and data protection concerns
 Interviewers‘ own behavior
 Interviewer as respondents
 Membership in social networks
 Income
 Blood donation
 Hypothetical questions:
disclose sensitive information
consent to record linkage
consent to biomarkers
Conceptual framework
 Interviewers’ experience with measurements
Conducting standardized interviews
SHARE
Conducting blood sugar tests
 Interviewers’ expectations of unit response,
consent and item response rates
Expected response and consent rates:
Different incentive groups
Biomarker measurements
Record linkage
Income
Some results of the interviewer survey
 Nonresponse
• Interested in learning about
the lives of other people
expect higher RRs
• Important to work on research
that is relevant to society
• Using of social networks
(facebook)
expect lower RRs
0
20
40
60
80
100
Expected unit response rates
no incentive
20 Euro
10 Euro
40 Euro
Interviewers were confident that the higher the value of the incentive the
more successful they would be in recruiting respondents.
Some results of the interviewer survey
 Consent to record linkage
Expected consent rate: 59,2%
Interviewers who would reveal personal information expect a
significantly higher consent rate.
SSN
Telephone number
Private Address
Address of health insurer
Some results of the interviewer survey
Interviewers who would consent to data linkage expect a
significantly higher consent rate.
Credit history
Employment history
Medical records
Social benefits
Interviewer who are part of social networks expect
significantly lower record linkage rates.
Outlook
 Next steps:
Completion of the survey
Linking with SHARE survey data to learn more about
interviewer effects
 Can we explain interviewer effects in SHARE with the
interviewers’ characteristics allocated in the interviewer
survey?