Jessica Herzing

Does the Recruitment of Offline Households Increase the Sample Representativeness of Probability-Based Online Panels? Evidence From the German Internet Panel

The past decade has seen a rise in the use of online panels for conducting survey research. However, the popularity of online panels, largely driven by relatively low implementation costs and high rates of Internet penetration, has been met with criticisms regarding their ability to accurately represent their intended target populations. This criticism largely stems from the fact that (1) non-Internet (or offline) households, despite their relatively small size, constitute a highly selective group unaccounted for in Internet panels, and (2) the preeminent use of nonprobability-based recruitment methods likely contributes a self-selection bias that further compromises the representativeness of online panels. In response to these criticisms, some online panel studies have taken steps to recruit probability-based samples of individuals and providing them with the means to participate online. Using data from one such study, the German Internet Panel, this article investigates the impact of including offline households in the sample on the representativeness of the panel. Consistent with studies in other countries, we find that the exclusion of offline households produces significant coverage biases in online panel surveys, and the inclusion of these households in the sample improves the representativeness of the survey despite their lower propensity to respond.

The Influence of a Person’s Digital Affinity on Unit Nonresponse and Attrition in an Online Panel:

Research has shown that the non-Internet population is hesitant to respond to online survey requests. However, also subgroups in the Internet population with low digital affinity may hesitate to respond to online surveys. This latter issue has not yet received much attention by scholars despite its potentially detrimental effects on the external validity of online survey data. In this article, we explore the extent to which a person’s digital affinity contributes to nonresponse bias in the German Internet Panel, a probability-based online panel of the general population. With a multidimensional classification of digital affinity, we predict response to the first online panel wave and participation across panel waves. We find that persons who belong to different classes of digital affinity have systematically different sociodemographic characteristics and show different voting behavior. In addition, we find that initial response propensities vary by classes of digital affinity, as do attrition patterns over time. Our results demonstrate the importance of digital affinity for the reduction in nonresponse bias during fieldwork and for postsurvey adjustments.

Face-to-face Rekrutierung für ein probabilistisches Onlinepanel

Onlineumfragen haben fur die sozialwissenschaftliche Forschung in den vergangen Jahren zunehmend an Bedeutung gewonnen, nicht zuletzt wegen der Moglichkeit, viele Personen innerhalb kurzer Zeit kosteneffizient befragen zu konnen. Im Vergleich zu interviewer-administrierten Befragungen haben Onlineinterviews aufgrund der selbst-administrierten Befragung einen geringeren Messfehler (Kreuter et al. 2008). Hauptkritikpunkt bei Onlineumfragen ist die Reprasentativitat dieser Studien im Hinblick auf die Allgemeinbevolkerung.