It has been acknowledged, that only probability-based online surveys can claim their results to be generalizable to the target population. In Germany a probability-based online access panel is organized by GESIS. Respondents are recruited by telephone (landline and mobile).

Such a recruitment strategy poses a problem in terms of non-response as there are several decisions to be made by potential panelists. They have to take part in the telephone interview, agree to become a member of the panel and fill in the online questionnaires. This multistep selection is a potential threat to the quality and generalizability of data collected online.

In this proposal we take a first step to assess the quality of the panel by comparing the data about online panelists with reference data about Internet users from the German General Social Survey (ALLBUS 2010).

We concentrate primarily on demographic characteristics, however some attitudinal measures are also considered. Respondents joining the panel may differ from ALLBUS Internet users since participation in the panel presupposes greater commitment opposed to one-time participation. Therefore, to assess the quality of the sample composition, the probability to participate in the online panel is studied for those who have agreed to a telephone interview. As a result, non-response error is estimated and the quality of the online panel can be assessed.