PresentationPDF Available

Explaining Interviewer Effects on Unit Nonresponse: A Cross-Survey Analysis.

Authors:
POLITICAL ECONOMY
OF REFORMS MAN NHE IM
SFB 884
Explaining Interviewer Effects on Unit
Nonresponse: A Cross-Survey Analysis
Annelies Blom 1 2, Daniela Ackermann-Piek 2,
Julie Korbmacher 3 4 and Ulrich Krieger 2
1School of Social Sciences, University of Mannheim
2German Internet Panel, SFB 884, University of Mannheim
3Munich Center for the Economics of Aging,
Max Planck Institute for Social Law and Social Policy
4Bayerisches Institut f¨ur Daten, Analysen und Qualit¨atssicherung
July 2017
1 / 15
Background
IInterviewer effects on nonresponse are well documented
IExplanations of those interviewer effects on nonresponse are
inconsistent across studies
ITo explain differences, comparable data needed
IInformation on interviewers
IField procedures
ISurvey organizations
2 / 15
Aim
IUsing four surveys with similar characteristics and data
collected
ISame country
ISimilar field period (2012-14)
ISame survey organization (TNS Infratest Sozialforschung -
Kantar Public)
ICall record data: contact and cooperation
IAuxiliary data
ISimilar interviewer survey data
IExamining interviewer effects on contact and cooperation
IKeeping as many factors constant as possible
IFind consistent interviewer effects on contact and cooperation
3 / 15
Studies
IGerman Internet Panel (GIP) recruitment interview 2012
IGerman Internet Panel (GIP) recruitment interview 2014
IProgramme for the International Assessment of Adult
Competencies (PIAAC) Germany
ISurvey of Health, Ageing and Retirement in Europe (SHARE)
Germany
4 / 15
Acknowledgements -I-
IPIAAC in Germany was funded by the Federal Ministry of
Education and Research (BMBF) with the participation of the
Federal Ministry of Labor and Social Affairs (BMAS).
IThe German Internet Panel is funded by the German Research
Foundation (DFG) as part of the Research Center (SFB) 884
Political Economy of Reforms
5 / 15
Acknowledgements -II-
IThe SHARE data collection has been primarily funded by the
European Commission through the FP5
(QLK6-CT-2001-00360), FP6 (SHARE-I3:
RII-CT-2006-062193, COMPARE: CIT5-CT-2005-028857,
SHARELIFE: CIT4-CT-2006-028812) and FP7
(SHARE-PREP: N211909, SHARE-LEAP: N227822, SHARE
M4: N261982). Additional funding from the German Ministry
of Education and Research, the U.S. National Institute on
Aging (U01 AG0974013S2, P01 AG005842, P01 AG08291,
P30 AG12815, R21 AG025169, Y1AG4553-01,
IAG BSR06-11, OGHA 04-064) and from various national
funding sources is gratefully acknowledged (see
www.share-project.org).
ISHARE W5 interviewer survey funded by the Charles Cannell
Fund in Survey Methodology
6 / 15
Method
IDependent variables contact and cooperation
ISeparate multilevel models (level: PSU)
INo random allocation of interviewers; therefore controlling for
area effects
IIncluding interviewer level variables in the model (interviewer
survey)
IData harmonized across studies
7 / 15
Descriptives
GIP 2012
GIP 2014
PIAAC
SHARE
Number of cases 4589 5942 7989 7175
Number of interviewers 131 141 115 142
Mean contact propensity 0.87 0.91 0.96 0.98
Mean cooperation propensity 0.43 0.54 0.61 0.35
8 / 15
Results - Controlling for area effects
Iarea effect variables: population size, geographical region,
unemplyment quota, proportion of single hh, proportion of
German citizens
Table: Intraclass correlation from multilevel regression on contact and
cooperation. Level variable: interviewer
GIP 2012
GIP 2014
PIAAC
SHARE
Contact
intercept only 26.4 16.9 20.0 57.1
including PSU variables 18.5 17.2 20.7 60.6
Cooperation
intercept only 14.6 8.3 2.1 6.8
including PSU variables 12.8 6.5 2.1 5.2
IA small percentage of variance in cooperation and contact
rates can be explained by including sampling point
information.
9 / 15
Multilevel logistic regression on contact
-5 -3 -1 1 3 5
GIP 2012
-5 -3 -1 1 3 5
GIP 2014
-5 -3 -1 1 3 5
PIAAC
-5 -3 -1 1 3 5
SHARE
46 to 60 years old
Older than 60
The Interviewer is female
Interviewer is educated
Interviewer is employed
5 to 10 years of experience
More than 10 years of experience
16 to 30 work hours
More than 30 work hours
Tailor content
Tailor questions to shorten survey
Tailor to adapt to respondent
Stick to instructions
Success via hard work
Acceptance of refusal
Respect towards voluntariness
Caught at right time, most will cooperate
Concerned about data protection
Science
People
Formal conditions
Age and gender of interviewer
Education and employment status
Work hours and experience of interviewer
Interviewers attitudes
Reasons for working as interviewer
y
10 / 15
Multilevel logistic regression on contact - summary
GIP 2012
GIP 2014
PIAAC
SHARE
Int. female +
Int. employed +
Int. working 16 to 30 hours -
Tailor questions to shorten survey -
Stick to instructions +
Respect towards voluntariness -
Controlling for area characteristics and ns interviewer effects.
Level variable: interviewer
11 / 15
Multilevel logistic regression on cooperation
-1.5 -0.5 0.5 1.5
GIP 2012
-1.5 -0.5 0.5 1.5
GIP 2014
-1.5 -0.5 0.5 1.5
SHARE
46 to 60 years old
Older than 60
The Interviewer is female
Interviewer is educated
Interviewer is employed
5 to 10 years of experience
More than 10 years of experience
16 to 30 work hours
More than 30 work hours
Tailor content
Tailor questions to shorten survey
Tailor to adapt to respondent
Stick to instructions
Success via hard work
Acceptance of refusal
Respect towards voluntariness
Caught at right time, most will cooperate
Concerned about data protection
Science
People
Formal conditions
Age and gender of interviewer
Education and employment status
Work hours and experience of interviewer
Interviewers attitudes
Reasons for working as interviewer
Y
12 / 15
Multilevel logistic regression on cooperation - summary
GIP 2012
GIP 2014
SHARE
Int. female +
Int. educated - -
Int.: 5 to 10 y. of experience +
Int: More than 10 y. of experience +
Tailor content +
Acceptance of refusal +
Most resp. will cooperate +
Controlling for area characteristics and ns interviewer effects.
Level variable: interviewer
13 / 15
Discussion
IAim: Explaining interviewer effects on contact and
cooperation
IEffects found in literature very diverse
ITherefore, our study keeps country, data collection period,
agency (interviewer pool), available data etc. constant
INonetheless, no consistent interviewer effects found
IRemaining factors of influence: type of study conducted
(topic, sponsor, research team)
14 / 15
Thank you!
ulrich.krieger@uni-mannheim.de
15 / 15
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.