Web surveys are becoming more common, not only due to their low cost. Unfortunately, the quality of web surveys may be problematic since coverage problems and non-response may be severe. Such problems do not automatically lead to considerably biased results, since the representativity of the data is most important. It is clear that a high response rate is still a good indicator for the representativity too. On the other hand, a response rate is rarely high in face-to-face or telephone surveys and consequently the representativity may be poor in these surveys as well. The paper compares the success of those three data collection modes as reliably as possible. Hence, our survey designs are as equal as possible. Our application area, crime victimization, is demanding also because the sensitiveness of questions varies. Our conclusions are not for these reasons straightforward.

Keywords: Face-to-face interviewing; Non-response vs. measurement error; Probit regression; Web survey

Biography: Specialist in survey methodology