How to make internet surveys representative: A case study of a two-step weighting procedure

Internet surveys are becoming increasingly popular. Concerns about the representativeness of online samples, however, frequently cast doubts on the validity of conclusions derived from internet survey data. These doubts rest on the fact that not all persons have internet access and on the premise that people are more likely to participate in a survey if the subject matter interests them. This paper presents a two-step procedure for weighting data from online surveys that is based on an explicit behavioral model of internet access and survey participation decisions. We illustrate the application of this scheme in a case study of Perspektive Deutschland, a very large online survey that was conducted in Germany in 2001 with almost 170,000 online respondents. We discuss practical aspects of weighting procedures such as problems caused by large weights as well as theoretical aspects such as the statistical implications of weighting.