Searching for best casi practices: experiments for a new data collection method

This paper provides the follow up to the paper reporting on an experiment with computer assisted self interviewing (CASI) by Bart Soels and Costa Tchaoussoglou in Florence (Soels/Tchaoussoglou, 1999). In the past three years, CASI has been tested as a data collection method for the Dutch National Readership Survey, with positive results. The majority of respondents is very well able to fill in the questionnaire by themselves, even if they lack prior experience with computers. In the few cases where a respondents is not able to fill in the questionnaire and the interviewer has to take over (CAPI), this does not influence the readership figures. Despite of this absence of interviewer bias – which is a major advantage in comparison to CATI – a contractor’s effect (differences in results between two or more research companies) is still be found, which is surprising, since the contractor’s effect is usually attributed to differences in the work of interviewers.
CASI can be used in face to face interviews as well as in a computer panel (multi channel distribution). Although there are differences in the results between the two, this is mainly due to differences of the sample composition. If the same kind of respondents are compared, the differences are very small or disappear. A representative sample is therefore as important as ever. When recruiting respondents for the face to face interviews, telephone screening proves to be superiour to a face to face approach.

Download Paper (PDF)


Symposium: ,
Authors: ,