The data quality of online access panels can be threatened by self-selection processes into the panel (nonresponse), coverage error and mode/ response effects. Self-selection and noncoverage is likely to lead to biased sample estimates, while mode effects and mode-specific response effects preclude any generalization of outcomes produced by one survey mode to another. To study these effects, we built up a large access panel for the adult population of Germany using probability sampling for the recruitment of people by phone (landline and cell phones). Possible access panel modes are landline, cell phone and web. The project is part of the Priority Programme 1292 on 'Survey Methodology’ (see www.survey-methodology.de).

In this paper, the data of the recruitment interview is used to study noncoverage and nonresponse on the way into the panel. Four steps can be distinguished preceding the participation in the online track of the access panel: Internet usage, readiness to participate in the panel generally, readiness to participate online (and give the email address as contact information) and panel participation. Logistic regression analysis is used to study the effect of sociodemographic variables, political interest, survey attitudes, meta-questions on the first interview and other attitude questions on each of the selection steps.

The results presented show that both noncoverage and nonresponse threaten the validity of inferences from online access panels. Possible causes of online-nonresponse are discussed and prospects for future research on web panels are given.