I am delighted to see Mirta Galesic's and Michael Bosnjak's 2003 study of the impact of questionnaire length on response quality in Web surveys finally make it into print, in this case the summer issue of POQ. We all believe down to our toes that long surveys are bad, but I have been disappointed on more than one occasion to see how poorly prepared we sometimes are to make the arguments against length to clients who too often want surveys that stretch beyond what we believe to be reasonable.
Mirta and Michael help to fill that gap in a well-designed experiment. I'm not going to describe the details of the design but just go right to key finding: As survey length and complexity increase response quality declines. Respondents spend less time answering individual questions, skip questions or choose non-substantive responses (like DK) more often, key fewer characters in open ends, show less response differentiation in grid-style questions, and are more likely to quit the survey completely. These behaviors become especially prevalent at around 20 minutes.
Many studies since, have replicated these findings. I expect few readers of the article will be surprised by what they see. Google 'questionnaire length and response quality' and you'll see what I mean.
The obvious question: since there is all of this empirical evidence why does survey length continue to be such a problem?
.