Bella Struminskaya (GESIS)
Online surveys are gaining in importance as a tool of data collection and as a platform for methodological experiments. However, certain challenges in regard to the quality of data need to be considered. Volunteer as well as probability-based online panels pose problems of non-coverage bias. Internet users differ from non-users in demographic characteristics: they are younger, better educated, non-immigrants, and have higher income compared to non-users.
In addition to the differences in demographics there are some "webographic" or attitudinal dissimilarities between the two groups: Internet-users tend to be more "socially content" - they are more trusting, have lots of people to draw on for support, and tend to think that others are generally fair (Lenhart et al. 2003). Thus, attitudes and values of Internet users may differ from those of non-users and data collected online would lack representativeness.
The goal of this explorative study is to evaluate the data quality of an online panel comparing SOEP-data and data from the offline-recruited probability-based online panel (GESIS). The analysis includes three steps.
First, using SOEP2008 data the differences between Internet users and non-users are studied.
The results replicate the finding of Lehnart et al. showing that German Internet users tend to be more socially content: after controlling for demographic differences Internet users significantly differ in their opinion that generally people are trustworthy and fair (though that doesn't hold true for readiness to help others).
Second, the quality of the online panel is assessed by a comparison of the panelists' answers to the SOEP-data (the recruitment for the online panel started in December 2010). In the third step it would then be possible to estimate the non-coverage bias of the online panel by contrasting data collected via the Internet to the SOEP-data of the general population (Internet users and non-users).