Previous month:
June 2006
Next month:
August 2006

Posts from July 2006

Blind Studies on the Web

This is a tough issue.  A client supplies us with an email list of customers, former customers, or prospects with whom he'd like to do research, but he wants it blind so that respondents are not influenced by knowing who is sponsoring the research.  The CASRO code currently requires that we identify in the survey invitation where we got the respondent's email address.  In addition, this is a good way to stay clear of spam regulations, both legal and those sometimes used by ISPs.  A modification has been proposed to CASRO that would allow blind studies by promising to disclose the sponsor to the respondent after the survey is complete, but it has not yet been acted upon.

But it could well be that the underlying premise is incorrect, or at least not justified by the research record.  Maybe the effects of identifying the sponsor are not as problematic as some might think.   For starters, the mail survey literature shows that identifying the survey sponsor generally has a positive impact on respondent cooperation. Identifying the sponsor tends to lend credibility to the request and therefore increases the likelihood that a respondent will complete.  So in this regard, it's a positive.  But does it impact how people respond?  Maybe not.  I have seen a little bit of research in the mode effects literature to suggest that this problem is more serious when an interviewer is involved, less serious in self-administration.  It may well be that respondents view an interviewer as an extension of the sponsor, and therefore may be less likely to say negative things about that sponsor.  When asked to compare the sponsoring company to competitors, there may be a tendency to be more positive about the sponsor.  In self-administration, this dynamic is less important.  This is particularly the case if the invitation includes a strong statement on confidentiality.

Like so many of these methodological issues we can't really know the answer without doing some experimental work.  And even then, what works for a survey on one topic may not work as well on others, suggesting that any differences that may exist in response patterns between sponsor-identified and blind studies may not be great enough to warrant a good deal of concern.