Tuesday 14th July      Wednesday 15th July      Thursday 16th July      Friday 17th July     

Friday 17th July, 11:00 - 12:30 Room: HT-102

Representativeness of Surveys Using Internet-based Data Collection
Convenor	Professor Michael Bosnjak (GESIS - Leibniz Institute for the Social Sciences )
Coordinator 1	Mr Ulrich Krieger (University of Mannheim)
Coordinator 2	Dr Tobias Enderle (GESIS - Leibniz Institute for the Social Sciences)

Session Details

Statistical theory is essentially based on random probability samples. However, Web surveys making use of convenience samples and volunteer access panels, where respondents self-select themselves into the sample, still dominate the landscape. Such recruitment strategies are attractive due to their low costs. Yet, recent years have seen increasing debates surrounding their quality in terms of representativeness. As a consequence, researchers across several countries are working towards Internet and mixed-mode panels based on probability samples.

The overall aim of this session is to provide a platform to present and to discuss recent research on the representativeness of surveys making use of Internet-based data collection modes (i.e., Web surveys, self-administered mobile surveys). The scope of this sessions encompasses both cross-sectional as well as panel-based surveys using Internet surveys as the sole data collection mode, or as one mode within a mixed mode context. Proposals may include, but are not limited to, the following topics:

* Representativeness concepts and corresponding indicators applicable to Internet-based and mixed-mode surveys
* Representativeness of Internet-based surveys and mixed-mode surveys of the general population compared to well-established modes
* Sample recruitment and refreshment strategies aimed at ensuring representativeness in (Internet-based and mixed-mode) access panels
* Effectiveness of various survey implementation measures and procedures aimed at ensuring representativeness (e.g., effectiveness of panel maintenance strategies, incentives, non-responder conversion, etc.)

Paper Details

1. Conditional vs. Unconditional incentives: Comparing the effect on sample composition in the recruitment of the German Internet Panel study.
Mr Ulrich Krieger (German Internet Panel, SFB 884, University of Mannheim)

Recruitment into the German Internet Panel consisted of various stages: the face-to-face household interview, mailed invitations to the online survey, reminder letters, a phone follow-up, and final mailed reminders. During the face-to-face phase we conducted an experiment with €5 unconditional (prepaid) vs. €10 conditional household (promised) incentives.
In this presentation we to compare characteristics, opinions, attitudes and interest in the survey topic of sample members from the 5€ and the 10€ households on two stages of the process: face-to-face survey participation and online panel registration.

2. Evolution of representativeness in an online probability panel
Dr Annette Scherpenzeel (Max Planck Institute for Social Law and Social Policy)
Dr Thomas Klausch (Utrecht University)
Dr Barry Schouten (Statistics Netherlands and Utrecht University)

The LISS panel is a large-scale online panel established in 2007 on the basis of a random probability sample from the Dutch population. The panel has been in place between from 2007 to 2014 and has been maintained to the highest methodological standards currently available. Nevertheless, the panel has experienced drop-out of respondents in the course of its life time. In this project, we evaluate by means of representativeness (R-) indicators to which degree this attrition was selectivity hampering representativeness over the lifetime of the panel.

3. Moderators of Survey Representativeness: A Meta-Analysis
Ms Carina Cornesse (German Internet Panel, SFB 884,University of Mannheim, Germany)
Professor Michael Bosnjak (GESIS - Leibniz Institute for the Social Sciences)

One major aim of any survey is to allow for inferences about certain properties of a pre-defined population. The degree to which sample estimates match population values is commonly termed ´degree of representativeness´. However, it is largely unclear which features of surveys explain the degree of representativeness. Using meta-analytic techniques, the overall goal of this paper is to estimate the extent to which survey design features (survey administration mode, sample type, number of survey waves) and the different representativeness concepts and their respective operationalizations (kind of representativeness indicator, number and type of auxiliary variables) explain a surveys´ representativeness.

4. Setting-up a probability-based web panel. Lessons learned from the ELIPSS Pilot Study.
Ms Anne-sophie Cousteaux (Sciences Po - CDSP)
Ms Anne Cornilleau (Sciences Po - CDSP)
Mr Stéphane Legleye (INED)

During the last decade, the internet has moved from an attractive data collection mode to a common way to administer surveys. The ELIPSS Panel (Étude longitudinale par internet pour les sciences sociales) is one of the probability-based web panels across Europe. This paper focuses on the lessons learned from the pilot study to set up the main panel. First, we will give an overview of the recruitment procedure planned for 2015 in the light of the pilot results. Then, we will present the specifications for the management of the panel to maintain monthly participation and limit attrition.

5. The pretest of Wave 2 of the German Health Interview and Examination Survey for Children and Adolescents (KiGGS) as a mixed-mode survey: composition of participant groups
Mr Robert Hoffmann (Department of Epidemiology and Health Monitoring, Robert Koch Institute, Berlin, Germany)
Mr Panagiotis Kamtsiuris (Department of Epidemiology and Health Monitoring, Robert Koch Institute, Berlin, Germany)

One aim of KiGGS is to collect representative data of children and adolescents and their parents. In the pretest of Wave 2, a mixed-mode survey design (paper/web offer) was evaluated regarding the net sample composition. The gross sample comprised 11,140 addresses which were randomly allocated to five experimental groups. The net samples were composed similarly regarding socio-demographic attributes. Among the participants who used the web mode, the fraction of highly educated participants was significantly higher than in the group who used paper questionnaires. Ultimately, KiGGS Wave 2 is conducted in a single-mode design.