In recent years, web surveys have incessantly become more popular. Part of this popularity can be explained by the low costs of web surveys, their high data collection speeds, and so on. However, it has frequently been noted that web surveys suffer from such major drawbacks as the absence of sampling frames, low coverage and low response rates. Here, we will focus on the sampling problem due to the lack of (complete) lists of Internet users and the inability to randomly generate e-mail addresses. One of the solutions to this problem is to construct a web survey panel and subsequently use that panel to draw samples from. We will discuss how web panels can be constructed and contrast this to face-to-face panels. However, most of our attention will be directed at panel mortality. Both in face-to-face panels and in web panels, certain panel members will cease to cooperate at some point in time. Our central research question is whether the reasons for panel attrition in both survey modes are similar and in which ways. Put differently, to what degree do the factors that influence panel attrition in face-to-face panels work in similar ways in web panels? Do the same factors (e.g. individualism; knowledge on the topic; interest; evaluation of the usefulness, agreeability and length of the survey; response behavior etc.) have the same explanatory power, or does changing the survey mode also fundamentally changes the reasons for dropping out of a panel? Our analyses are based on a three-wave face-to-face panel survey (conducted in 1991, 1995, and 1998) and a three-wave web panel survey (conducted in 2002-2003) on roughly the same topics and using roughly the same predictors.

presentation Loosveldt