Comparison of three survey modes

The paper compares the success of face‐to‐face, phone and web data collection modes. The
survey designs are as identical as possible. Our application area, crime victimisation, is
demanding in many ways, not least because the sensitiveness of the questions varies. Our
research questions, respectively, are twofold. First, we assess the mode effects from the point
of view of their frame coverage on one hand, and non‐response, on the other. We exploit this
analysis for developing ideal adjusted sample weights in order to compare differences in
measurement. In the second phase, our standpoints are that the mode that secures better the
secrecy for the respondent will produce higher victimisation prevalence, and the differences
between the modes are larger when more sensitive questions are asked. Our results show that
sensitive questions were not working well in telephone interviewing. It seems that the
representativity of surveys depends on several factors, not for instance only on a high
response rate.