ESRA Conference—Probability Panels Have Arrived in Europe in a Big Way
Academicians are very sensitive to the sample frame. Market researchers should be as well but we have adopted the non-probabilistic access panels as our main online sample source. It was refreshing to see how many talks were devoted to probabilistically created panels. For us it was a sampling heaven.
Based on the talks we heard, the Europeans are still pioneering on their use of probabilistic panels. No large commercial panels of the type have made their appearance and those that exist are academic or government based. They are still testing the waters but are clearly sensitive to the issues.
A battery of studies came out of work performed by the CentERdata Tilburg University, the Netherlands. In 2007 CentERdata started the LISS panel (Longitudinal Internet Studies for the Social Sciences), which is “an online panel which is based on a true probability sample of households.” It is representative of the Dutch population and consists of 5000 members. This may be small by commercial standards but apparently they are producing quite a bit of research as we were privileged to listen to three good presentations.
The LISS panel began with a random sample of some 8000 respondents in 2007. Those who did not have internet connection were provided with a computer and a connection. By 2010 approximately 5000 were still active (see www.lissdata.nl). The “attriters” were found to be different from the “stayers” in that the latter were more conscientious, open and female. Those that tend to leave the panel were found to be Christian Democrats less often.
“Lurkers”, those who come and go in terms of completing some or many of the 25 monthly waves of the subject study, tended to have a lower level of education.
Those who abandoned the panel entirely, as opposed to the lurkers who came and went, were labeled “monotonic attriters”.
This is the first paper that we know of, besides our own, that drew a behavioral distinction between attriters and stayers. We consider this to be an important finding; One that we do not feel that we can do justice to here. But the concept of a changing panel population created by differential attrition is one that we must begin to address.
Interesting in that this comment appears to regard the UK as not part of Europe. One of the biggest and most successful non-probability panels (in Europe) is YouGov’s in the UK. There are also several other more market research oriented panels of a similar kind. LISS is interesting but hardly representative. In the UK the dominant mode among probability sampling preferring researchers working on panel studies is now *conversion* of existing face to face panels into at least partial on line mode for later waves.